BOOLE-1.1 notes

general algorithm:

  1. compute the intersection curve between two solids
  2. merge the intersection curves within each patch
  3. merge curves between patches in each of the two solids
  4. traverse each solid's topological graph to generate "partitions" using the merged intersection curves
  5. compute IN/OUT for each partition and apply this state to each patch in that partition
  6. for each component that will be included in the new solid, stitch the trim patches together
  7. construct a graph of connectivity/adjacency between components/partitions to keep track of the "face/patch map", etc.

perform_CSG algorithm:

  1. CalculateImplicit for each solid -- convert from the parameterized nurbs or bezier surface to the implicit representation. bernSteinToPowerBasis(tpPatch0) //tensor patch 0 implicitize(tpPatch0, ndet) //paper explains this, it's just some fancy substitution moveFactoredToFinalDeterminant return nDetImplicit, lame pointer stuff

  2. findXsectionCurve for each of the two given solids - returns intersection curves for each patch in the solid, compute bounding boxes find all overlapping bounding boxes magic goes here to further reduce the number of possible intersections to check for each patch pair, doLoopDetection if loop, HandleLoopCase else, GetInsidePointsUV Param_1_in_Imp_0 //complicated spline merging goes here, OR Merge2DLists to just use a linear piecewise approximation of the intersection curve Now project each curve of the second patch boundary on the domain of the first patch. (use ProjectPath) return_st_end_in_p2d // returns the start and end points of a merged curve FillXscArray Now project each curve of the first patch boundary on the domain of the second patch.. (repeat)

  3. Merge intersection curves further

  4. CheckIntersectionCurves -- check that the curves in each solid form a closed loop on the surface, if not die immediately. After this it gets funky?

  5. PartitionSplinegon for each trim portion in each tensor patch on each solid, then set AdjInfo after calling GetAdjacentPartn (for the adjacency graph later)

  6. For every patch of secnod_solid that overlaps with the first_solid on the same_side, assign portions of the boundary that actually lie inside the first solid the same_solid value 0. Inside of this loop, call AdjustTrimCurves.

  7. Now split some trim curves if the same_solid flag is 0. This is because the curve being cut on the other side belongs to the other solid, so the curve must be an intersection curve. If I split an intersection curve, I must do it on both sides to get proper adjacencies. This calls SubdivideTrimCurve and then GetAdjacentPartn.

SubdivideTrimCurve - takes a UVspline and a XYZspline and splits them into (n+1) curves given n param values

next steps for first_solid:

8. ObtainNewGlobalGraph - takes the new partitioned regions of a solid and assigns each region a unique global component number.
9. DetermineXsectionsforSolid - classifies each component of a solid as IN/OUT

next steps for second_solid:

8. ObtainNewGlobalGraph - takes the new partitioned regions of a solid and assigns each region a unique global component number.
9. DetermineXsectionsforSolid - classifies each component of a solid as IN/OUT

Then continue with these steps:

  1. ObtainFinalPiecesofSolid(first_solid)
  2. ObtainFinalPiecesofSolid(second_solid)
  3. SplitIntoNewRegions(second_solid)

then copy lots of data around

then free lots of memory

then destroy the moon.