Add normal= for 2D shapes

[sergei9838]

Hi, Marco,
It would be nice to be able to create 2D shapes with a given orientation like it is now with Plane, so that the shape is on the plane with a given normal vector. It does not require much: just to re-use your transformation code in Plane. For instance, Circle class may look like this:

from typing import List, Union, Any
class Circle(Polygon):
    """
    Build a Circle of radius `r`.
    """
    def __init__(self, pos=(0, 0, 0), r=1.0, normal=(0, 0, 1), res=120, c="gray5", alpha=1.0) -> None:
        """
        Build a Circle of radius `r` in the plane orthogonal to `normal` and passing through `pos`.
        """
        super().__init__(pos=(0, 0, 0), nsides=res, r=r)
        nl = np.asarray(normal, dtype=float)
        axis = nl / np.linalg.norm(nl)
        theta = np.arccos(axis[2])
        phi = np.arctan2(axis[1], axis[0])

        self.transform = LinearTransform()
        self.transform.rotate_y(np.rad2deg(theta))
        self.transform.rotate_z(np.rad2deg(phi))
        self.transform.translate(pos)
        self.apply_transform(self.transform)

        self.nr_of_points = 0
        self.va = 0
        self.vb = 0
        self.axis1: List[float] = []
        self.axis2: List[float] = []
        self.center: List[float] = []  # filled by pointcloud.pca_ellipse()
        self.pvalue = 0.0  # filled by pointcloud.pca_ellipse()
        self.alpha(alpha).c(c)
        self.name = "Circle"

    def acircularity(self) -> float:
        """
        Return a measure of how different an ellipse is from a circle.
        Values close to zero correspond to a circular object.
        """
        a, b = self.va, self.vb
        value = 0.0
        if a + b:
            value = ((a - b) / (a + b)) ** 2
        return value

I added 10 lines after super(), argument normal and property transform

Usage:

oriented_circle = Circle(normal=(1,1,1))
oriented_circle.show(axes=1).close()
print(oriented_circle.transform)

Not to repeat the same code, perhaps, it would be better to have a function like align_with_normal which would apply the above transform

If you think it might break many codes, Oriented_ subclasses can be created:

class Oriented_circle(Circle):
    def __init__(self, pos=(0, 0, 0), r=1.0, normal=(0, 0, 1), res=120, c='gray5', alpha=1.0):
        super().__init__((0, 0, 0), r, res, c, alpha)
        nl = np.asarray(normal, dtype=float)
        axis = nl / np.linalg.norm(nl)
        theta = np.arccos(axis[2])
        phi = np.arctan2(axis[1], axis[0])

        self.transform = LinearTransform()
        self.transform.rotate_y(np.rad2deg(theta))
        self.transform.rotate_z(np.rad2deg(phi))
        self.transform.translate(pos)
        self.apply_transform(self.transform)

Hopefully, it will be useful

Cheers,
Sergei

[marcomusy]

Hi Sergei, in principle this is an already available functionality, except one needs to use it after instantiation of the object.
All vedo objects already have a transform field. For ex.:

from vedo import *

circle = Circle()
print(circle.transform)

circle.reorient([0,0,1], [1,1,1])
print(circle.transform)

show(circle, axes=1)

[sergei9838]

Oh, I didn't know about reorient - thank you so much! Worth an example... )

[marcomusy]

In fact the reorient() is a bit hacky and dirty... because it works ok for symmetric objects but may not yield a desired result for asymmetric one: for those you need to fully specify a 3 rotations sequence in space with .rotate_x(), etc..