Regression fix for Elements.Geometry.Solids.Sweep

[DmytroMuravskyi]

BACKGROUND:

• Where was reported an issue about Sweep: Regression Elements.Geometry.Solids.Sweep #1007
• The reason code worked before is that setbacks parameters were just ignored.
• Sweep in the issue crashes because setback ends exactly on second vertex.
• Setbacks need extra handling for special cases as it always assumes that start setback is on first segment, end - on last. This is not the case as they can easily exceed.

DESCRIPTION:

• Changed GetSubdivisionParameters to use startSetbackDistance and endSetbackDistance since they were just ignored.
• Changed Polyline.Frames and Polygon.Frames so they return only frames between setbacks. Since CreateMiterTransform and CreateOrthogonalTransform are focused on vertices, Transforms on setbacks are created manually.

TESTING:

• Added tests for updated functions as well as regression Sweep.

FUTURE WORK:

• Polyline.TransformAt feels incorrect as it says that u is "between 0.0 and length" but it's treated as 0 to 1 value. Also, since calculation inside is length based, including normal calculation, curve length can possibly influence frame vectors in a wrong way.

REQUIRED:

• All changes are up to date in CHANGELOG.md.

COMMENTS:

• I will be away for a while, so feel free to steal this code if necessery.

---

This change is 

[andrewheumann]

Reviewed 3 of 9 files at r1, all commit messages.
Reviewable status: 1 change requests, 0 of 1 approvals obtained (waiting on @DmytroMuravskyi)

---

Elements/src/Geometry/Polygon.cs line 2094 at r1 (raw file):

            // Calculate number of frames. 2 frames corresponding to end parameters.
            // 1 if startIndex == endIndex.
            var lenght = endIndex - startIndex + 3;

lenght => length

---

Elements/src/Geometry/Polyline.cs line 306 at r1 (raw file):

            // Calculate number of frames. 2 frames corresponding to end parameters.
            // 1 if startIndex == endIndex.
            var lenght =  endIndex - startIndex + 3;

lenght => length

---

Elements/test/PolygonTests.cs line 2234 at r1 (raw file):

            Assert.True((Vector3.YAxis - Vector3.XAxis).IsParallelTo(frames[4].ZAxis));

            frames = polygon.Frames(2, 2);

polyhon.Frames(1, 1) will throw an exception here.

[DmytroMuravskyi]

Reviewable status: 1 change requests, 0 of 1 approvals obtained (waiting on @andrewheumann)

---

Elements/src/Geometry/Polygon.cs line 2094 at r1 (raw file):

Previously, andrewheumann (Andrew Heumann) wrote…

lenght => length

Done.

---

Elements/src/Geometry/Polyline.cs line 306 at r1 (raw file):

Previously, andrewheumann (Andrew Heumann) wrote…

lenght => length

Done.

---

Elements/test/PolygonTests.cs line 2234 at r1 (raw file):

Previously, andrewheumann (Andrew Heumann) wrote…

polyhon.Frames(1, 1) will throw an exception here.

Good catch. Added index rounding for Polygon.

[andrewheumann]

The frames being found along the curve are flipped compared to how they used to be — I think this is an unintended regression. This also results in sweeps having "inside-out" normals:

Elements 1.6.0:

Elements 2.0.0:

This PR:

All three samples are running this code:

var polygon = new Polyline((0, 0), (3, 0), (3, 3), (0, 3));
var model = new Model();
model.AddElement(new ModelCurve(polygon, BuiltInMaterials.XAxis));
var frames = polygon.Frames(0.1, 0.1);
model.AddElements(frames.SelectMany(f => f.ToModelCurves()));

var profile = Polygon.L(0.2, 0.2, 0.05);
var ge = new GeometricElement {
    Representation = new Sweep(profile, polygon, 0.1, 0.1, 0, false)
};
model.AddElement(ge);

We have been inconsistent about this in the past, but I think the convention in our code is that the z of a curve's frame is opposite the curve's tangent at that point. (Don't ask me why it is this way, I would 100% expect Z+ to be the same as the tangent.)

Reviewed 3 of 9 files at r1, 3 of 3 files at r2, all commit messages.
Reviewable status: 1 change requests, 0 of 1 approvals obtained

[DmytroMuravskyi]

Yes, I missed the fact that they are adjusted. Now frames look like this

Note that Frames behaves differently from TrasformAt as first and last frames are not rotated based on position if they are somewhere in the middle of segment but just . One exception is when setback is coincides with a corner, then Frame is also the same as for corner vertex - rotated. I also made tests more explicit about expected results.

Reviewable status: 1 change requests, 0 of 1 approvals obtained

[andrewheumann]

Looks good to me, thanks!

[andrewheumann]

Reviewed 4 of 4 files at r3, all commit messages.
Reviewable status: complete! 1 of 1 approvals obtained