Optimize distance calculations (#203)

- use `numpy.linalg.norm` instead of `scipy.spatial.distance` (and remove scipy dependency)
- cache all `Circle` instances when we create an instance of `Rings`, and compute all separations via numpy

Co-authored-by: Stefanie Molin <24376333+stefmolin@users.noreply.github.com>

docs/conf.py

@@ -63,7 +63,6 @@
     'Pillow': ('https://pillow.readthedocs.io/en/stable/', None),
     'pytest': ('https://pytest.org/en/stable/', None),
     'python': ('https://docs.python.org/3/', None),
-    'scipy': ('https://docs.scipy.org/doc/scipy/', None),
 }
 
 

pyproject.toml

@@ -47,7 +47,6 @@ dependencies = [
   "matplotlib>=3.3",
   "numpy>=1.20",
   "pandas>=1.2",
-  "scipy>=1.10.0",
   "tqdm>=4.64.1",
 ]
 optional-dependencies.dev = [

src/data_morph/shapes/bases/line_collection.py

@@ -4,6 +4,7 @@
 from typing import Iterable
 
 import matplotlib.pyplot as plt
+import numpy as np
 from matplotlib.axes import Axes
 
 from ...plotting.style import plot_with_custom_style
@@ -75,8 +76,9 @@ def _distance_point_to_line(
 
         .. _this VBA code: http://local.wasp.uwa.edu.au/~pbourke/geometry/pointline/source.vba
         """
-        start, end = line
+        start, end = np.array(line)
         line_mag = self._euclidean_distance(start, end)
+        point = np.array(point)
 
         if line_mag < 0.00000001:
             # Arbitrarily large value
@@ -100,7 +102,7 @@ def _distance_point_to_line(
             # Intersecting point is on the line, use the formula
             ix = x1 + u * (x2 - x1)
             iy = y1 + u * (y2 - y1)
-            distance = self._euclidean_distance(point, (ix, iy))
+            distance = self._euclidean_distance(point, np.array((ix, iy)))
 
         return distance
 

src/data_morph/shapes/bases/shape.py

@@ -4,8 +4,8 @@
 from numbers import Number
 from typing import Iterable, Optional
 
+import numpy as np
 from matplotlib.axes import Axes
-from scipy.spatial import distance
 
 
 class Shape(ABC):
@@ -67,9 +67,9 @@ def _euclidean_distance(a: Iterable[Number], b: Iterable[Number]) -> float:
 
         See Also
         --------
-        scipy.spatial.distance.euclidean : Euclidean distance calculation.
+        numpy.linalg.norm : Euclidean distance calculation.
         """
-        return distance.euclidean(a, b)
+        return np.linalg.norm(a - b)
 
     def _recursive_repr(self, attr: Optional[str] = None) -> str:
         """

src/data_morph/shapes/circles.py

@@ -43,6 +43,8 @@ def __init__(self, dataset: Dataset, r: Number = None) -> None:
         self.r: Number = r or dataset.df.std().mean() * 1.5
         """numbers.Number: The radius of the circle."""
 
+        self._center = np.array([self.cx, self.cy])
+
     def __repr__(self) -> str:
         return f'<{self.__class__.__name__} cx={self.cx} cy={self.cy} r={self.r}>'
 
@@ -60,7 +62,7 @@ def distance(self, x: Number, y: Number) -> float:
         float
             The absolute distance between this circle's edge and the point (x, y).
         """
-        return abs(self._euclidean_distance((self.cx, self.cy), (x, y)) - self.r)
+        return abs(self._euclidean_distance(self._center, np.array([x, y])) - self.r)
 
     @plot_with_custom_style
     def plot(self, ax: Axes = None) -> Axes:
@@ -125,6 +127,9 @@ def __init__(self, dataset: Dataset, num_rings: int = 4) -> None:
         ]
         """list[Circle]: The individual rings represented by :class:`Circle` objects."""
 
+        self._centers = np.array([circle._center for circle in self.circles])
+        self._radii = np.array([circle.r for circle in self.circles])
+
     def __repr__(self) -> str:
         return self._recursive_repr('circles')
 
@@ -150,7 +155,10 @@ def distance(self, x: Number, y: Number) -> float:
             Rings consists of multiple circles, so we use the minimum
             distance to one of the circles.
         """
-        return min(circle.distance(x, y) for circle in self.circles)
+        point = np.array([x, y])
+        return np.min(
+            np.abs(np.linalg.norm(self._centers - point, axis=1) - self._radii)
+        )
 
     @plot_with_custom_style
     def plot(self, ax: Axes = None) -> Axes: