Feature/delaunay primitive

.github/workflows/main.yml

@@ -40,13 +40,6 @@ jobs:
         pip install ./PyAutoArray
         pip install ./PyAutoArray[optional]
 
-        cd PyAutoArray/autoarray/util/nn/src/nn
-        export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/home/runner/work/PyAutoArray/PyAutoArray/PyAutoArray/autoarray/util/nn/src/nn
-        bash ./configure
-        cp makefile_autolens makefile
-        make
-        cd /home/runner/work/PyAutoArray/PyAutoArray
-
     - name: Extract branch name
       shell: bash
       run: |
@@ -75,7 +68,6 @@ jobs:
         export ROOT_DIR=`pwd`
         export PYTHONPATH=$PYTHONPATH:$ROOT_DIR/PyAutoConf
         export PYTHONPATH=$PYTHONPATH:$ROOT_DIR/PyAutoArray
-        export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/home/runner/work/PyAutoArray/PyAutoArray/PyAutoArray/autoarray/util/nn/src/nn
         pushd PyAutoArray
         python3 -m pytest --cov autoarray --cov-report xml:coverage.xml
     - name: Slack send

.gitignore

@@ -33,6 +33,3 @@ test_autoarray/unit/structures/files/array.fits
 test_autoarray/unit/structures/files/frame.fits
 test_autoarray/unit/structures/files/kernel.fits
 test_autoarray/unit/structures/grids/files/grid/grid.fits
-
-# Ignoring all files in autoarray/util/nn/src/.
-autoarray/util/nn/src/*

autoarray/__init__.py

@@ -42,7 +42,6 @@
 from .inversion.pixelization.mappers.factory import mapper_from as Mapper
 from .inversion.pixelization.mappers.rectangular import MapperRectangular
 from .inversion.pixelization.mappers.delaunay import MapperDelaunay
-from .inversion.pixelization.mappers.voronoi import MapperVoronoi
 from .inversion.pixelization.mappers.rectangular_uniform import MapperRectangularUniform
 from .inversion.pixelization.image_mesh.abstract import AbstractImageMesh
 from .inversion.pixelization.mesh.abstract import AbstractMesh
@@ -77,7 +76,6 @@
 from .structures.grids.irregular_2d import Grid2DIrregular
 from .structures.mesh.rectangular_2d import Mesh2DRectangular
 from .structures.mesh.rectangular_2d_uniform import Mesh2DRectangularUniform
-from .structures.mesh.voronoi_2d import Mesh2DVoronoi
 from .structures.mesh.delaunay_2d import Mesh2DDelaunay
 from .structures.arrays.kernel_2d import Kernel2D
 from .structures.vectors.uniform import VectorYX2D

autoarray/abstract_ndarray.py

@@ -73,15 +73,6 @@ def __init__(self, array, xp=np):
         while isinstance(array, AbstractNDArray):
             array = array.array
         self._array = array
-        # try:
-        #     register_pytree_node(
-        #         type(self),
-        #         self.instance_flatten,
-        #         self.instance_unflatten,
-        #     )
-        # except ValueError:
-        #     pass
-
         self._xp = xp
 
     def invert(self):

autoarray/config/general.yaml

@@ -14,7 +14,5 @@ numba:
   cache: true
   nopython: true
   parallel: false
-pixelization:
-  voronoi_nn_max_interpolation_neighbors: 300
 structures:
   native_binned_only: false           # If True, data structures are only stored in their native and binned format. This is used to reduce memory usage in autocti.

autoarray/config/visualize/mat_wrap_2d.yaml

@@ -155,15 +155,6 @@ DelaunayDrawer:       # wrapper for `plt.fill()`: customize the appearance of De
     alpha: 0.7
     edgecolor: k
     linewidth: 0.0  
-VoronoiDrawer:       # wrapper for `plt.fill()`: customize the appearance of Voronoi mesh's.
-  figure:
-    alpha: 0.7
-    edgecolor: k
-    linewidth: 0.3
-  subplot:
-    alpha: 0.7
-    edgecolor: k
-    linewidth: 0.3
 ParallelOverscanPlot:
   figure:
     c: k

autoarray/fixtures.py

@@ -378,28 +378,9 @@ def make_delaunay_mesh_grid_9():
         pixel_scales=1.0,
     )
 
-    return aa.Mesh2DDelaunay(values=grid_9)
-
-
-def make_voronoi_mesh_grid_9():
-    grid_9 = aa.Grid2D.no_mask(
-        values=[
-            [0.6, -0.3],
-            [0.5, -0.8],
-            [0.2, 0.1],
-            [0.0, 0.5],
-            [-0.3, -0.8],
-            [-0.6, -0.5],
-            [-0.4, -1.1],
-            [-1.2, 0.8],
-            [-1.5, 0.9],
-        ],
-        shape_native=(3, 3),
-        pixel_scales=1.0,
-    )
-
-    return aa.Mesh2DVoronoi(
+    return aa.Mesh2DDelaunay(
         values=grid_9,
+        source_plane_data_grid_over_sampled=make_grid_2d_sub_2_7x7().over_sampled,
     )
 
 
@@ -445,22 +426,6 @@ def make_delaunay_mapper_9_3x3():
     )
 
 
-def make_voronoi_mapper_9_3x3():
-    mapper_grids = aa.MapperGrids(
-        mask=make_mask_2d_7x7(),
-        source_plane_data_grid=make_grid_2d_sub_2_7x7(),
-        source_plane_mesh_grid=make_voronoi_mesh_grid_9(),
-        image_plane_mesh_grid=aa.Grid2D.uniform(shape_native=(3, 3), pixel_scales=0.1),
-        adapt_data=aa.Array2D.ones(shape_native=(3, 3), pixel_scales=0.1),
-    )
-
-    return aa.MapperVoronoi(
-        mapper_grids=mapper_grids,
-        border_relocator=make_border_relocator_2d_7x7(),
-        regularization=make_regularization_constant(),
-    )
-
-
 def make_rectangular_inversion_7x7_3x3():
     return aa.Inversion(
         dataset=make_masked_imaging_7x7(),
@@ -475,12 +440,6 @@ def make_delaunay_inversion_9_3x3():
     )
 
 
-def make_voronoi_inversion_9_3x3():
-    return aa.Inversion(
-        dataset=make_masked_imaging_7x7(), linear_obj_list=[make_voronoi_mapper_9_3x3()]
-    )
-
-
 ### EUCLID DATA ####
 
 

autoarray/inversion/inversion/abstract.py

@@ -200,33 +200,6 @@ def all_linear_obj_have_regularization(self) -> bool:
             list(filter(None, self.regularization_list))
         )
 
-    @property
-    def mapper_edge_pixel_list(self) -> List[int]:
-        """
-        Returns the edge pixels of all mappers in the inversion.
-
-        This uses the `edge_pixel_list` property of the `Mesh` of the `Mapper` class, and updates their values to
-        correspond to the indexing of the overall inversion's `curvature_matrix`.
-
-        This is used to regulareze the edge pixels of the inversion's `reconstruction` or remove them from the
-        inversion procedure entirely (e.g. make these values of these edge pixels zero).
-
-        Returns
-        -------
-        A list of the edge pixels of all mappers in the inversion, where the values are updated to correspond to the
-        indexing of the overall inversion's `curvature_matrix`.
-        """
-        mapper_edge_pixel_list = []
-
-        param_range_list = self.param_range_list_from(cls=LinearObj)
-
-        for param_range, linear_obj in zip(param_range_list, self.linear_obj_list):
-            if isinstance(linear_obj, AbstractMapper):
-                for edge_pixel in linear_obj.edge_pixel_list:
-                    mapper_edge_pixel_list.append(edge_pixel + param_range[0])
-
-        return mapper_edge_pixel_list
-
     @property
     def total_regularizations(self) -> int:
         return sum(

autoarray/inversion/inversion/mapper_valued.py

@@ -16,7 +16,7 @@ def __init__(self, mapper, values, mesh_pixel_mask: Optional[np.ndarray] = None)
         mapper pixel) in order to perform calculations which use both the `Mapper` and these values.
 
         For example, a common use case is to interpolate the reconstruction of values on a mapper from the
-        mesh of the mapper (e.g. a Voronoi mesh) to a uniform Cartesian grid of values, because the irregular mesh
+        mesh of the mapper (e.g. a Delaunay mesh) to a uniform Cartesian grid of values, because the irregular mesh
         is difficult to plot and analyze.
 
         This class also provides functionality to compute the magnification of the reconstruction, by comparing the
@@ -26,7 +26,7 @@ def __init__(self, mapper, values, mesh_pixel_mask: Optional[np.ndarray] = None)
         Parameters
         ----------
         mapper
-            The `Mapper` object which pairs with the values, for example a `MapperVoronoi` object.
+            The `Mapper` object which pairs with the values, for example a `MapperDelaunay` object.
         values
             The values of each pixel of the mapper, which could be the `reconstruction` values of an `Inversion`,
             but alternatively could be other quantities such as the noise-map of these values.
@@ -71,7 +71,7 @@ def interpolated_array_from(
         extent: Optional[Tuple[float, float, float, float]] = None,
     ) -> Array2D:
         """
-        The values of a mapper can be on an irregular pixelization (e.g. a Delaunay triangulation, Voronoi mesh).
+        The values of a mapper can be on an irregular pixelization (e.g. a Delaunay triangulation).
 
         Analysing the reconstruction can therefore be difficult and require specific functionality tailored to using
         this irregular grid.
@@ -166,7 +166,7 @@ def max_pixel_centre(self) -> Grid2DIrregular:
         max_pixel = np.argmax(self.values_masked)
 
         max_pixel_centre = Grid2DIrregular(
-            values=[self.mapper.source_plane_mesh_grid[max_pixel]]
+            values=[self.mapper.source_plane_mesh_grid.array[max_pixel]]
         )
 
         return max_pixel_centre
@@ -218,7 +218,8 @@ def magnification_via_mesh_from(
         PSF, as the source plane reconstruction is a non-convolved image.
 
         In the source-plane, this is computed by summing the reconstruction values multiplied by the area of each
-        mesh pixel, for example if the source-plane is a `Voronoi` mesh this is the area of each Voronoi pixel.
+        mesh pixel, for example if the source-plane is a `Delaunay` mesh this is the area of each corresponding
+        Delaunay pixel.
 
         This calculatiion is generally more robust that using an interpolated
         image (see `magnification_via_interpolation_from`), because it uses the exact the source-plane reconstruction
@@ -244,7 +245,7 @@ def magnification_via_mesh_from(
 
                 To compute the magnification of a `Delaunay` mesh, use the method `magnification_via_interpolation_from`.
 
-                This method only supports a `RectangularMagnification` or `Voronoi` mesh.
+                This method only supports a `RectangularMagnification`.
                 """
             )
 
@@ -257,8 +258,7 @@ def magnification_via_mesh_from(
                 """
                 The magnification cannot be computed because the areas of the source-plane mesh pixels are all zero.
 
-                This probably means you have specified an invalid source-plane mesh, for example a `Voronoi` mesh
-                where all pixels are on the edge of the source-plane and therefore have an infinite border.
+                This probably means you have specified an invalid source-plane mesh.
                 """
             )
 

autoarray/inversion/linear_obj/neighbors.py

@@ -4,8 +4,7 @@
 class Neighbors(np.ndarray):
     def __new__(cls, arr: np.ndarray, sizes: np.ndarray):
         """
-        Class packaging ndarrays describing the neighbors of every pixel in a mesh (e.g. `RectangularMagnification`,
-        `Voronoi`).
+        Class packaging ndarrays describing the neighbors of every pixel in a mesh (e.g. `RectangularMagnification`).
 
         The array `arr` contains the pixel indexes of the neighbors of every pixel. Its has shape [total_pixels,
         max_neighbors_in_single_pixel].
@@ -28,7 +27,7 @@ def __new__(cls, arr: np.ndarray, sizes: np.ndarray):
 
         - For pixel 4, the central pixel, neighbors[4,:] = [1, 3, 5, 7] and neighbors_sizes[4] = 4.
 
-        The same arrays can be generalized for other pixelizations, for example a `Voronoi` grid.
+        The same arrays can be generalized for other pixelizations, for example a `Delaunay` grid.
 
         Parameters
         ----------

autoarray/inversion/mock/mock_mapper.py

@@ -14,10 +14,9 @@ def __init__(
         over_sampler=None,
         border_relocator=None,
         adapt_data=None,
-        edge_pixel_list=None,
         regularization=None,
         pix_sub_weights=None,
-        pix_sub_weights_split_cross=None,
+        pix_sub_weights_split_points=None,
         mapping_matrix=None,
         pixel_signals=None,
         parameters=None,
@@ -37,9 +36,8 @@ def __init__(
         )
 
         self._over_sampler = over_sampler
-        self._edge_pixel_list = edge_pixel_list
         self._pix_sub_weights = pix_sub_weights
-        self._pix_sub_weights_split_cross = pix_sub_weights_split_cross
+        self._pix_sub_weights_split_points = pix_sub_weights_split_points
         self._mapping_matrix = mapping_matrix
         self._parameters = parameters
         self._pixel_signals = pixel_signals
@@ -62,17 +60,13 @@ def over_sampler(self):
             return super().over_sampler
         return self._over_sampler
 
-    @property
-    def edge_pixel_list(self):
-        return self._edge_pixel_list
-
     @property
     def pix_sub_weights(self):
         return self._pix_sub_weights
 
     @property
-    def pix_sub_weights_split_cross(self):
-        return self._pix_sub_weights_split_cross
+    def pix_sub_weights_split_points(self):
+        return self._pix_sub_weights_split_points
 
     @property
     def mapping_matrix(self):

autoarray/inversion/pixelization/border_relocator.py

@@ -412,7 +412,7 @@ def relocated_mesh_grid_from(
 
             relocated_grid = relocated_grid_from(
                 grid=mesh_grid.array,
-                border_grid=grid[self.border_slim],
+                border_grid=grid[self.sub_border_slim],
                 xp=xp,
             )
 

autoarray/inversion/pixelization/image_mesh/overlay.py

@@ -201,7 +201,6 @@ def image_plane_mesh_grid_from(
         adapt_data
             Not used by this image mesh.
         """
-
         pixel_scales = mask.pixel_scales
 
         grid = mask.derive_grid.unmasked

autoarray/inversion/pixelization/mappers/abstract.py

@@ -25,6 +25,7 @@ def __init__(
         mapper_grids: MapperGrids,
         regularization: Optional[AbstractRegularization],
         border_relocator: BorderRelocator,
+        preloads=None,
         xp=np,
     ):
         """
@@ -88,6 +89,7 @@ def __init__(
 
         self.border_relocator = border_relocator
         self.mapper_grids = mapper_grids
+        self.preloads = preloads
 
     @property
     def params(self) -> int:
@@ -113,10 +115,6 @@ def image_plane_mesh_grid(self) -> Grid2D:
     def over_sampler(self):
         return self.mapper_grids.source_plane_data_grid.over_sampler
 
-    @property
-    def edge_pixel_list(self) -> List[int]:
-        return self.source_plane_mesh_grid.edge_pixel_list
-
     @property
     def adapt_data(self) -> np.ndarray:
         return self.mapper_grids.adapt_data
@@ -420,7 +418,7 @@ def interpolated_array_from(
     ) -> Array2D:
         """
         The reconstructed values of a mapper (e.g. the `reconstruction` of an `Inversion` may be on an irregular
-        pixelization (e.g. a Delaunay triangulation, Voronoi mesh).
+        pixelization (e.g. a Delaunay triangulation).
 
         Analysing the reconstruction can therefore be difficult and require specific functionality tailored to using
         this irregular grid.
@@ -437,7 +435,7 @@ def interpolated_array_from(
         ----------
         values
             The value corresponding to the reconstructed value of every pixelization pixel (e.g. Delaunay triangle
-            vertexes, Voronoi mesh cells).
+            vertexes).
         shape_native
             The 2D shape in pixels of the interpolated reconstruction, which is always returned using square pixels.
         extent