Revision: 73931
http://sourceforge.net/p/brlcad/code/73931
Author: starseeker
Date: 2019-09-16 12:33:42 +0000 (Mon, 16 Sep 2019)
Log Message:
-----------
Stash some preliminary thoughts on how to better find a surface point from the
trim - other issues to deal with first, but I want to come back to this.
Modified Paths:
--------------
brlcad/trunk/src/libbrep/cdt_edge.cpp
Modified: brlcad/trunk/src/libbrep/cdt_edge.cpp
===================================================================
--- brlcad/trunk/src/libbrep/cdt_edge.cpp 2019-09-14 21:32:44 UTC (rev
73930)
+++ brlcad/trunk/src/libbrep/cdt_edge.cpp 2019-09-16 12:33:42 UTC (rev
73931)
@@ -128,6 +128,71 @@
}
#endif
+#if 0
+// If we have an associated 3D edge, we need a surface point that will
+// result in a sensible triangle near that edge. Construct and evaluate
+// the proposed 3D triangle, and use that to evaluate the suitability of
+// the proposed surface point
+bool
+candidate_surf_pnt(struct ON_Brep_CDT_State *s_cdt, cdt_mesh::cpolyedge_t *pe)
+{
+ if (!pe->eseg) return false;
+
+ // TODO - find the smallest angle between this segment and the prev/next
+ // segs - that may (will?) constrain how far out into the surface we can
+ // safely go. If we're unconstrained try for equilateral, else bisect the
+ // angle and intersect that vector with the vector off the midpoint. (i.e.
+ // use trig to get the length to which to scale ndir)
+ //
+ // If we're unconstrained, we'll still need to check the candidate point
+ // against the rtree to see if we're closer to another edge elsewhere in
+ // the loop than we our to pe. if we are, shorten the distance until we
+ // either make the triangle too narrow or we get an unusable triangle in
+ // 3D. if either of those happen, this edge can't supply a trim point.
+ //
+ // This has the potential to allow for sparser minimal splitting than the
+ // box overlap test, since as long as we can produce a usable associated
+ // surface point we are OK to use a trim without further splitting.
+ //
+ // Note the candidate must also be closer to the midpoint than either of
+ // the neighboring segment points - if that distance is shorter than the
+ // equilateral distance, it's our start just as an angle constraint limits
+ // our start...
+
+
+ // TODO - do we ever need to flip these?
+ ON_3dPoint *p1 = pe->eseg->e_start;
+ ON_3dPoint *p2 = pe->eseg->e_end;
+
+ ON_BrepTrim& trim = s_cdt->brep->m_T[pe->trim_ind];
+ ON_2dPoint p2d1(pe->polygon->pnts_2d[pe->v[0]].first,
pe->polygon->pnts_2d[pe->v[0]].second);
+ ON_2dPoint p2d2(pe->polygon->pnts_2d[pe->v[1]].first,
pe->polygon->pnts_2d[pe->v[1]].second);
+
+ ON_3dPoint p13d(p2d1.x, p2d1.y, 0);
+ ON_3dPoint p23d(p2d2.x, p2d2.y, 0);
+ ON_3dPoint p1norm(p2d1.x, p2d1.y, 1);
+ ON_3dVector vtrim = p23d - p13d;
+ ON_3dVector vnorm = p1norm - p13d;
+ vtrim.Unitize();
+ vnorm.Unitize();
+ ON_3dVector ndir = ON_CrossProduct(vnorm, vtrim);
+ ndir.Unitize();
+
+
+ double dist = p2d1.DistanceTo(p2d2);
+ double bdist = 0.5*dist;
+ ndir = ndir * bdist;
+ ON_2dVector ndir2d(ndir.x, ndir.y);
+ ON_2dPoint p2mid = (p2d1 + p2d2) * 0.5;
+ pe->spnt = p2mid + ndir2d;
+ pe->defines_spnt = true;
+
+
+}
+#endif
+
+
+
// TODO - can we base the width of this box on how far the 3D edge curve
midpoint is from
// the midpoint of the 3D line segment? That might relate the 2D box size to
how far away
// we need to be to avoid problematic surface points...
This was sent by the SourceForge.net collaborative development platform, the
world's largest Open Source development site.
_______________________________________________
BRL-CAD Source Commits mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/brlcad-commits
