Fix network region recognition.

[ikeough]

BACKGROUND:
While creating networks of walls in Hypar, I attempted to recognize closed areas as rooms by using Newtork.FindAllClosedRegions. I found for the shape pictured that the red cell would not be found. This was because the traversal algorithm was previously visiting all the nodes in such a way as to ignore 4 and 5 as having been previously traversed, resulting in an unclosed region.

DESCRIPTION:
This PR adds a new traversal method, Network.TraverseLeftWithoutLeaves, and uses this traversal method in the Network.FindAllClosedRegions method. This new traversal method ignores leaf nodes, both external to the graph, and internal to a cell, resulting in less traversals of unclosed paths. It reliably "turns left" resulting in proper closed regions.

TESTING:

• See the NetworkTests.FourQuadrantsWithPointerFindsCorrectNumberOfClosedRegions test.

FUTURE WORK:

• Network region recognition still does not work on rooms with a concave shape.

REQUIRED:

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

---

This change is 

[ikeough]

Fixes #900.

[andrewheumann]

Works great except for a couple cases, which can be left for future work.

You've already identified that Concave shapes are an issue — this could be just another manifestation of the same problem, but this will fail to find both regions:

var lines = new List<Line> {
    new Line((10.00, 0.00), (10.00, 10.00)),
    new Line((10.00, 10.00), (-0.00, 10.00)),
    new Line((-0.00, 10.00), (0.00, 0.00)),
    new Line((0.00, 0.00), (10.00, 0.00)),
    new Line((-2.00, 6.00), (12.00, 6.00)),
    new Line((5.00, -2.00), (4.00, 1.00)),
    new Line((4.00, 1.00), (6.00, 2.00)),
    new Line((6.00, 2.00), (5.00, 4.00)),
};

This one misses a region, returning 4 instead of 5, even though they should all be convex:

var lines = new List<Line> {
    new Line((10.00, 0.00), (10.00, 10.00)),
    new Line((10.00, 10.00), (-0.00, 10.00)),
    new Line((-0.00, 10.00), (0.00, 0.00)),
    new Line((0.00, 0.00), (10.00, 0.00)),
    new Line((-2.00, 6.00), (12.00, 6.00)),
    new Line((-2.96, -0.16), (2.13, 2.51)),
    new Line((2.13, 2.51), (7.81, 2.51)),
    new Line((7.81, 2.51), (13.11, 0.28)),
    new Line((7.81, 2.51), (11.60, 4.88)),
    new Line((2.13, 2.51), (-1.46, 4.62)),
};

Reviewed 1 of 2 files at r1, 2 of 2 files at r2, all commit messages.
Reviewable status: 1 change requests, 0 of 1 approvals obtained (waiting on @ikeough and @katehryhorenko)

---

Elements/src/Search/Network.cs line 555 at r2 (raw file):

        /// <summary>
        /// Traverse a network following the left-most candidate edge relative to the current edge.
        /// leaf nodes.

this comment was maybe unfinished?

---

Elements/src/Search/Network.cs line 559 at r2 (raw file):

        /// <param name="traversalData">Data about the current step of the traversal.</param>
        /// <param name="allNodeLocations">A collection of all node locations in the network.</param>
        /// <param name="visitedEdges">A collection of previously visited edge.</param>

previously visited edges

---

Elements/src/Search/Network.cs line 593 at r2 (raw file):

                var localEdge = (allNodeLocations[e] - allNodeLocations[traversalData.currentIndex]).Unitized();
                var angle = 180 - baseEdge.PlaneAngleTo(localEdge);

Could be future work, but this embeds an assumption that the network is 2D and in the XY plane. Can we at least leave a comment w/ a TODO to accept a normal or otherwise support non-planar networks?

[andrewheumann]

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

---

Elements/src/Search/Network.cs line 498 at r2 (raw file):

        /// Traverse a network following the smallest plane angle between the current
        /// edge and the next candidate edge.
        /// leaf nodes.

unfinished comment?

[ikeough]

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

---

Elements/src/Search/Network.cs line 498 at r2 (raw file):

Previously, andrewheumann wrote…

unfinished comment?

Done.

---

Elements/src/Search/Network.cs line 555 at r2 (raw file):

Previously, andrewheumann wrote…

this comment was maybe unfinished?

Done.

---

Elements/src/Search/Network.cs line 559 at r2 (raw file):

Previously, andrewheumann wrote…

previously visited edges

Done.

---

Elements/src/Search/Network.cs line 593 at r2 (raw file):

Previously, andrewheumann wrote…

Could be future work, but this embeds an assumption that the network is 2D and in the XY plane. Can we at least leave a comment w/ a TODO to accept a normal or otherwise support non-planar networks?

I've added a clarification to the Network.FromSegmentableItems method's docs. That's where the assumption is embedded first as the algorithm uses a line sweep on planar line segments to construct the network.

[ikeough]

@andrewheumann

[andrewheumann]

Reviewed 1 of 2 files at r3, 1 of 1 files at r4, all commit messages.
Reviewable status: complete! 1 of 1 approvals obtained