Merge pull request #401 from jcapriot/tensor_p2i

Add point2index functionality for `tensor_mesh`

Author: jcapriot

discretize/tensor_mesh.py

@@ -6,7 +6,7 @@
 from discretize.base import BaseRectangularMesh, BaseTensorMesh
 from discretize.operators import DiffOperators, InnerProducts
 from discretize.mixins import InterfaceMixins, TensorMeshIO
-from discretize.utils import mkvc
+from discretize.utils import mkvc, as_array_n_by_dim
 from discretize.utils.code_utils import deprecate_property
 
 from .tensor_cell import TensorCell
@@ -756,6 +756,33 @@ def cell_boundary_indices(self):
             indzu = self.gridCC[:, 2] == max(self.gridCC[:, 2])
             return indxd, indxu, indyd, indyu, indzd, indzu
 
+    def point2index(self, locs):  # NOQA D102
+        # Documentation inherited from discretize.base.BaseMesh
+
+        locs = as_array_n_by_dim(locs, self.dim)
+        # in each dimension do a sorted search within the nodes
+        # arrays to find the containing cell in that dimension
+        cell_bounds = [
+            self.nodes_x,
+        ]
+        if self.dim > 1:
+            cell_bounds.append(self.nodes_y)
+        if self.dim == 3:
+            cell_bounds.append(self.nodes_z)
+
+        # subtract 1 here because given the nodes [0, 1], the point 0.5 would be inserted
+        # at index 1 to maintain the sorted list, but that corresponds to cell 0.
+        # clipping here ensures that anything outside the mesh will return the nearest cell.
+        multi_inds = tuple(
+            np.clip(np.searchsorted(n, p) - 1, 0, len(n) - 2)
+            for n, p in zip(cell_bounds, locs.T)
+        )
+        # and of course, we are fortran ordered in a tensor mesh.
+        if self.dim == 1:
+            return multi_inds[0]
+        else:
+            return np.ravel_multi_index(multi_inds, self.shape_cells, order="F")
+
     def _repr_attributes(self):
         """Represent attributes of the mesh."""
         attrs = {}

tests/base/test_tensor.py

@@ -1,5 +1,6 @@
 import pytest
 import numpy as np
+import numpy.testing as npt
 import unittest
 import discretize
 from scipy.sparse.linalg import spsolve
@@ -321,5 +322,70 @@ def test_orderBackward(self):
         self.orderTest()
 
 
+@pytest.fixture(params=[1, 2, 3], ids=["dims-1", "dims-2", "dims-3"])
+def random_tensor_mesh(request):
+    dim = request.param
+    rng = np.random.default_rng(440122)
+    shape = rng.integers(5, 10, dim)
+    cell_widths = [rng.uniform(3.0, 872634.321, n) for n in shape]
+    origin = rng.uniform(-101.031, 33.2, dim)
+
+    return discretize.TensorMesh(cell_widths, origin)
+
+
+def test_tensor_point2index_inside_points(random_tensor_mesh):
+    mesh = random_tensor_mesh
+    dim = mesh.dim
+    m_origin = mesh.origin
+    m_extent = np.atleast_1d(np.max(mesh.nodes, axis=0))
+
+    nd = 15
+    points = np.stack(np.meshgrid(*np.linspace(m_origin, m_extent, nd).T), axis=-1)
+    points = points.reshape((-1, dim))
+
+    npt.assert_array_equal(mesh.is_inside(points), True)
+
+    cell_inds = mesh.point2index(points)
+    for icell, p in zip(cell_inds, points):
+        cell = mesh[icell]
+        c_origin, c_extent = cell.bounds.reshape((dim, 2)).T
+        dim_test = (p >= c_origin) & (p <= c_extent)
+        npt.assert_equal(dim_test, True)
+
+
+def test_tensor_point2index_outside_points(random_tensor_mesh):
+    mesh = random_tensor_mesh
+    dim = mesh.dim
+    m_origin = mesh.origin
+    m_extent = np.atleast_1d(np.max(mesh.nodes, axis=0))
+    m_width = m_extent - m_origin
+
+    nd = 15
+    points = np.stack(
+        np.meshgrid(*np.linspace(m_origin - m_width * 2, m_extent + m_width * 2, nd).T),
+        axis=-1,
+    )
+    points = points.reshape((-1, dim))
+    outside_points = points[~mesh.is_inside(points)]
+
+    npt.assert_array_equal(mesh.is_inside(outside_points), False)
+
+    # manually check each point that is outside
+    cell_inds = mesh.point2index(outside_points)
+    for icell, p in zip(cell_inds, outside_points):
+        cell = mesh[icell]
+        c_origin, c_extent = cell.bounds.reshape((dim, 2)).T
+        dim_test = np.zeros(dim, bool)
+        for i in range(dim):
+            p_d = p[i]
+            if p_d < m_origin[i]:
+                dim_test[i] = p_d < c_origin[i]
+            elif p_d > m_extent[i]:
+                dim_test[i] = p_d > c_extent[i]
+            else:
+                dim_test[i] = p_d >= c_origin[i] and p_d <= c_extent[i]
+        npt.assert_equal(dim_test, True)
+
+
 if __name__ == "__main__":
     unittest.main()