fix Ge.Matrix3d.scale3d()

[gswifort]

>>> Ge.Matrix3d.scaling(-5, Ge.Point3d.kOrigin).scale3d()
PyGe.Scale3d(5.00000000000000,5.00000000000000,5.000000000000000)  # -5.0 != 5.0

I think this PR solves this problem.

[Copilot]

Pull request overview

This PR fixes a bug in Matrix3d.scale3d() where negative scaling factors were being returned as positive values. The issue occurred because the method used length() to extract scale from coordinate system vectors, which always returns positive values. The fix extracts the diagonal components directly (x.x, y.y, z.z) to preserve the sign of scaling factors.

Changes:

• Modified AcGeMatrix3dGetScaling3d() to use x.x, y.y, z.z instead of x.length(), y.length(), z.length() when returning the Scale3d object

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

The 2D version of this function (AcGeMatrix2dGetScaling2d at line 865) still uses x.length() and y.length(), which would have the same issue with negative scaling factors. Consider applying the same fix to the 2D version for consistency: return AcGeScale2d(x.x, y.y) instead of AcGeScale2d(x.length(), y.length()).

[gswifort]

@copilot open a new pull request to apply changes based on this feedback

[Copilot]

Consider adding a test case for negative scaling to verify this fix works correctly. For example, test that Matrix3d.scaling(-5, Point3d.kOrigin).scale3d() returns Scale3d(-5, -5, -5) and not Scale3d(5, 5, 5).

[gswifort]

@copilot open a new pull request to apply changes based on this feedback

[Copilot]

Pull request overview

Copilot reviewed 1 out of 1 changed files in this pull request and generated 1 comment.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

This change fixes negative scaling for axis-aligned matrices but breaks behavior for non-axis-aligned transformations.

For axis-aligned scaling (e.g., Matrix3d.scaling(-5, ...)):

• Basis vectors: x=(-5,0,0), y=(0,-5,0), z=(0,0,-5)
• Old: (5, 5, 5) - incorrect, loses sign ✗
• New: (-5, -5, -5) - correct ✓

For rotated transformations (e.g., 45° rotation, no scale):

• Basis vectors: x≈(0.707,0.707,0), y≈(-0.707,0.707,0), z=(0,0,1)
• Old: (1, 1, 1) - correct magnitudes ✓
• New: (0.707, 0.707, 1) - incorrect, not meaningful scale values ✗

Consider one of these approaches:

1. If scale3d() is only intended for axis-aligned matrices, add documentation stating this limitation
2. Use a proper decomposition: sign(determinant) * length() for uniform scale, or sign(x.x) * x.length() for each axis where sign() uses the component with largest absolute value
3. Check if vectors are axis-aligned (only one non-zero component) and use component if true, else use length()

[CEXT-Dan]

hmm, yeah, length is clearly wrong.

though what if I rotate the matrix?

consider

        xf = Ge.Matrix3d.kIdentity
        xf.setToRotation(math.radians(45),Ge.Vector3d.kZAxis,Ge.Point3d.kOrigin)
        print(xf.xAxis().x, xf.yAxis().y,xf.zAxis().z)
        print(xf.scale3d())

[CEXT-Dan]

this seems to work

def extract_scale(xf: Ge.Matrix3d):
    vecs = [xf.xAxis(), xf.yAxis(), xf.zAxis()]
    orig = [Ge.Vector3d.kXAxis, Ge.Vector3d.kYAxis, Ge.Vector3d.kZAxis]
    scales = []
    for i in range(3):
        length = vecs[i].length()
        if vecs[i].dotProduct(orig[i]) < 0:
            scales.append(-length)
        else:
            scales.append(length)
    return tuple(scales)

def test():
    xf = Ge.Matrix3d.kIdentity
    pl = Ge.Plane(Ge.Point3d.kOrigin, Ge.Vector3d.kXAxis)
    xf.setToMirroring(pl)
    print(extract_scale(xf)) #(1.0, 1.0, 1.0) wrong
 
    xf = Ge.Matrix3d.kIdentity
    pl = Ge.Plane(Ge.Point3d.kOrigin, Ge.Vector3d.kYAxis)
    xf.setToMirroring(pl)
    print(extract_scale(xf)) #(1.0, 1.0, 1.0) wrong
    
    xf = Ge.Matrix3d.kIdentity
    pl = Ge.Plane(Ge.Point3d.kOrigin, Ge.Vector3d.kZAxis)
    xf.setToMirroring(pl)
    print(extract_scale(xf)) #(1.0, 1.0, 1.0) wrong
    
    xf = Ge.Matrix3d.kIdentity
    xf.setToRotation(math.radians(45), Ge.Vector3d.kZAxis, Ge.Point3d.kOrigin)
    print(extract_scale(xf)) #(1.0, 1.0, 1.0) correct