Distance structs TODO item

[nsmela]

Had some spare time, tried to handle a simple TODO item.

Classes within /distance converted to struct, aside from static Distance. Considering adding an Obsolete attribute to it?

MeshQueries hit a snag: it returns null on lines 17, 65, and 123. Structs cannot be null and I think wrapping with nullable defeats purpose of switching to struct. For now, returning a new struct.

Taking a guess on what you'd prefer to in the future for MeshQueries:

• returns ResultOrFail
• out parameter: public static void TriangleDistance(DMesh3 mesh, int ti, Vector3d point, out DistPoint3Triangle3 value)
• Create an empty version of each struct
• output (bool success, DistPoint3Triangle3 result)
• int ti for triangle index becomes a struct that must be valid to be made. The mesh it is coming from would create it. This would move the failure closer to the start of the operation:

public struct TriangleId(int id);

public ResultOrFail<TriangleId> DMesh3.GetTriangleId(int index);

[nsmela]

Ran some tests, found a bug. It may have existed before:

Test_DistPoint2Circle2
Message: 

Expected: 5.0d +/- 9.9999999999999995E-07d

But was: 7.0710678118654755d

Off by: -2.0710678118654755d

Gemini's explanation:

In geometry3Sharp, adding a scalar (double) to a Vector2d adds that scalar to both components ($x+s, y+s$).
For a radius of 5, this resulted in (0,0) + 5 = (5,5).
The distance from center (0,0) to (5,5) is $\sqrt{5^2 + 5^2} \approx 7.071$, causing the test failure.

CircleClosest = circle.Center;
CircleClosest.x += circle.Radius;

[rms80]

yes that's a bug, make you can make a separate PR for it as I can't accept this entire thing right now.

This is a major breaking change in the library API, because something that used to be a by-reference type is changed to a by-value type. IMO doing it w/ structs is better, but a lot of code is written assuming they are classes, and changing to by-value potentially could introduce subtle bugs in library user code (eg if someone is passing the object to a function that initializes it, that wouldn't work w/ a struct unless it's the argument is changed to be passed as ref, etc).

The TODO was basically a note to myself to spend some time thinking about this. IE do I want to say that the new 'dotnet8' version is not backwards-compatible w/ the old version, and fix mistakes like this, or be safe and do things like make the struct versions have new names and keep/deprecate the class versions, etc. Also there is a bunch of internal usage in the library of these types that needs to be investigated / cleaned-up.

So, appreciate the effort but I don't have time to think about all this right now, sorry. I will keep the PR open in case I get motivated on some weekend.

[CLAassistant]

All committers have signed the CLA.

[nsmela]

Understood!