Unidata/MetPy

View on GitHub
src/metpy/interpolate/geometry.py

Summary

Maintainability
A
0 mins
Test Coverage
# Copyright (c) 2018 MetPy Developers.
# Distributed under the terms of the BSD 3-Clause License.
# SPDX-License-Identifier: BSD-3-Clause
"""Tools for working with geometric objects (points, triangles, polygons)."""

import logging
import math

import numpy as np
from scipy.spatial import KDTree

log = logging.getLogger(__name__)


def get_points_within_r(center_points, target_points, r):
    r"""Get all target_points within a specified radius of a center point.

    All data must be in same coordinate system, or you will get undetermined results.

    Parameters
    ----------
    center_points: (X, Y) numpy.ndarray
        location from which to grab surrounding points within r
    target_points: (X, Y) numpy.ndarray
        points from which to return if they are within r of center_points
    r: integer
        search radius around center_points to grab target_points

    Returns
    -------
    matches: (X, Y) numpy.ndarray
        A list of points within r distance of, and in the same
        order as, center_points

    """
    tree = KDTree(target_points)
    indices = tree.query_ball_point(center_points, r)
    return tree.data[indices].T


def get_point_count_within_r(center_points, target_points, r):
    r"""Get count of target points within a specified radius from center points.

    All data must be in same coordinate system, or you will get undetermined results.

    Parameters
    ----------
    center_points: (X, Y) numpy.ndarray
        locations from which to grab surrounding points within r
    target_points: (X, Y) numpy.ndarray
        points from which to return if they are within r of center_points
    r: integer
        search radius around center_points to grab target_points

    Returns
    -------
    matches: (N, ) numpy.ndarray
        A list of point counts within r distance of, and in the same
        order as, center_points

    """
    tree = KDTree(target_points)
    indices = tree.query_ball_point(center_points, r)
    return np.array([len(x) for x in indices])


def triangle_area(pt1, pt2, pt3):
    r"""Return the area of a triangle.

    Parameters
    ----------
    pt1: (X,Y) numpy.ndarray
        Starting vertex of a triangle
    pt2: (X,Y) numpy.ndarray
        Second vertex of a triangle
    pt3: (X,Y) numpy.ndarray
        Ending vertex of a triangle

    Returns
    -------
    area: float
        Area of the given triangle.

    """
    a = 0.0

    a += pt1[0] * pt2[1] - pt2[0] * pt1[1]
    a += pt2[0] * pt3[1] - pt3[0] * pt2[1]
    a += pt3[0] * pt1[1] - pt1[0] * pt3[1]

    return abs(a) / 2


def dist_2(x0, y0, x1, y1):
    r"""Return the squared distance between two points.

    This is faster than calculating distance but should
    only be used with comparable ratios.

    Parameters
    ----------
    x0: float
        Starting x coordinate
    y0: float
        Starting y coordinate
    x1: float
        Ending x coordinate
    y1: float
        Ending y coordinate

    Returns
    -------
    d2: float
        squared distance

    See Also
    --------
    distance

    """
    d0 = x1 - x0
    d1 = y1 - y0
    return d0**2 + d1**2


def distance(p0, p1):
    r"""Return the distance between two points.

    Parameters
    ----------
    p0: (X,Y) numpy.ndarray
        Starting coordinate
    p1: (X,Y) numpy.ndarray
        Ending coordinate

    Returns
    -------
    d: float
        distance

    See Also
    --------
    dist_2

    """
    return math.sqrt(dist_2(p0[0], p0[1], p1[0], p1[1]))


def circumcircle_radius(pt0, pt1, pt2):
    r"""Calculate and return the radius of a given triangle's circumcircle.

    Parameters
    ----------
    pt0: (x, y)
        Starting vertex of triangle
    pt1: (x, y)
        Second vertex of triangle
    pt2: (x, y)
        Final vertex of a triangle

    Returns
    -------
    r: float
        circumcircle radius

    See Also
    --------
    circumcenter

    """
    a = distance(pt0, pt1)
    b = distance(pt1, pt2)
    c = distance(pt2, pt0)

    t_area = triangle_area(pt0, pt1, pt2)

    return (a * b * c) / (4 * t_area) if t_area > 0 else np.nan


def circumcenter(pt0, pt1, pt2):
    r"""Calculate and return the circumcenter of a circumcircle generated by a given triangle.

    All three points must be unique or a division by zero error will be raised.

    Parameters
    ----------
    pt0: (x, y)
        Starting vertex of triangle
    pt1: (x, y)
        Second vertex of triangle
    pt2: (x, y)
        Final vertex of a triangle

    Returns
    -------
    cc: (x, y)
        circumcenter coordinates

    See Also
    --------
    circumcenter

    """
    a_x = pt0[0]
    a_y = pt0[1]
    b_x = pt1[0]
    b_y = pt1[1]
    c_x = pt2[0]
    c_y = pt2[1]

    bc_y_diff = b_y - c_y
    ca_y_diff = c_y - a_y
    ab_y_diff = a_y - b_y
    cb_x_diff = c_x - b_x
    ac_x_diff = a_x - c_x
    ba_x_diff = b_x - a_x

    d_div = (a_x * bc_y_diff + b_x * ca_y_diff + c_x * ab_y_diff)

    if d_div == 0:
        raise ZeroDivisionError

    d_inv = 0.5 / d_div

    a_mag = a_x**2 + a_y**2
    b_mag = b_x**2 + b_y**2
    c_mag = c_x**2 + c_y**2

    cx = (a_mag * bc_y_diff + b_mag * ca_y_diff + c_mag * ab_y_diff) * d_inv
    cy = (a_mag * cb_x_diff + b_mag * ac_x_diff + c_mag * ba_x_diff) * d_inv

    return cx, cy


def find_natural_neighbors(tri, grid_points):
    r"""Return the natural neighbor triangles for each given grid cell.

    These are determined by the properties of the given Delaunay triangulation.
    A triangle is a natural neighbor of a grid cell if that triangle's circumcenter
    is within the circumradius of the grid cell center.

    Parameters
    ----------
    tri: `scipy.spatial.Delaunay`
        A Delaunay Triangulation.
    grid_points: (X, Y) numpy.ndarray
        Locations of grids.

    Returns
    -------
    members: dict
        List of simplex codes for natural neighbor triangles in ``tri`` for each grid cell.
    circumcenters: numpy.ndarray
        Circumcenter for each triangle in ``tri``.

    """
    # Used for fast identification of points with a radius of another point
    tree = KDTree(grid_points)

    # Mask for points that are outside the triangulation
    in_triangulation = tri.find_simplex(tree.data) >= 0

    circumcenters = []
    members = {key: [] for key in range(len(tree.data))}
    for i, indices in enumerate(tri.simplices):
        # Find the circumcircle (center and radius) for the triangle.
        triangle = tri.points[indices]
        cc = circumcenter(*triangle)
        r = circumcircle_radius(*triangle)
        circumcenters.append(cc)

        # Find all grid points within the circumcircle.
        for point in tree.query_ball_point(cc, r):
            # If this point is within the triangulation, add this triangle to its list of
            # natural neighbors
            if in_triangulation[point]:
                members[point].append(i)

    return members, np.array(circumcenters)


def find_nn_triangles_point(tri, cur_tri, point):
    r"""Return the natural neighbors of a triangle containing a point.

    This is based on the provided Delaunay Triangulation.

    Parameters
    ----------
    tri: `scipy.spatial.Delaunay`
        A Delaunay Triangulation
    cur_tri: int
        Simplex code for Delaunay Triangulation lookup of
        a given triangle that contains 'position'.
    point: (x, y)
        Coordinates used to calculate distances to
        simplexes in 'tri'.

    Returns
    -------
    nn: (N, ) array
        List of simplex codes for natural neighbor
        triangles in 'tri'.

    """
    nn = []

    candidates = set(tri.neighbors[cur_tri])

    # find the union of the two sets
    candidates |= set(tri.neighbors[tri.neighbors[cur_tri]].flat)

    # remove instances of the "no neighbors" code
    candidates.discard(-1)

    for neighbor in candidates:

        triangle = tri.points[tri.simplices[neighbor]]
        cur_x, cur_y = circumcenter(triangle[0], triangle[1], triangle[2])
        r = circumcircle_radius(triangle[0], triangle[1], triangle[2])

        if dist_2(point[0], point[1], cur_x, cur_y) < r**2:

            nn.append(neighbor)

    return nn


def find_local_boundary(tri, triangles):
    r"""Find and return the outside edges of a collection of natural neighbor triangles.

    There is no guarantee that this boundary is convex, so ConvexHull is not
    sufficient in some situations.

    Parameters
    ----------
    tri: `scipy.spatial.Delaunay`
        A Delaunay Triangulation
    triangles: (N, ) array
        List of natural neighbor triangles.

    Returns
    -------
    edges: list
        List of vertex codes that form outer edges of
        a group of natural neighbor triangles.

    """
    edges = []

    for triangle in triangles:

        for i in range(3):

            pt1 = tri.simplices[triangle][i]
            pt2 = tri.simplices[triangle][(i + 1) % 3]

            if (pt1, pt2) in edges:
                edges.remove((pt1, pt2))

            elif (pt2, pt1) in edges:
                edges.remove((pt2, pt1))

            else:
                edges.append((pt1, pt2))

    return edges


def area(poly):
    r"""Find the area of a given polygon using the shoelace algorithm.

    Parameters
    ----------
    poly: (2, N) numpy.ndarray
        2-dimensional coordinates representing an ordered
        traversal around the edge a polygon.

    Returns
    -------
    area: float

    """
    a = 0.0
    n = len(poly)

    for i in range(n):
        a += poly[i][0] * poly[(i + 1) % n][1] - poly[(i + 1) % n][0] * poly[i][1]

    return abs(a) / 2.0


def order_edges(edges):
    r"""Return an ordered traversal of the edges of a two-dimensional polygon.

    Parameters
    ----------
    edges: (2, N) numpy.ndarray
        List of unordered line segments, where each
        line segment is represented by two unique
        vertex codes.

    Returns
    -------
    ordered_edges: (2, N) numpy.ndarray

    """
    edge = edges[0]
    edges = edges[1:]

    ordered_edges = [edge]

    num_max = len(edges)
    while len(edges) > 0 and num_max > 0:

        match = edge[1]

        for search_edge in edges:
            vertex = search_edge[0]
            if match == vertex:
                edge = search_edge
                edges.remove(edge)
                ordered_edges.append(search_edge)
                break
        num_max -= 1

    return ordered_edges