Purpose: incremental rerouting during interaction (MR-8): when a
node drags, its incident edges must (a) stay anchored at the node
BORDER, leaving perpendicular to the attach side, and (b) never
route over other nodes. ELK owns from-scratch scene routing; this
router owns the interactive path.
Method: the classic sparse orthogonal visibility grid with A*
search and a bend penalty, per the published literature on
object-avoiding orthogonal connector routing (Wybrow, Marriott &
Stuckey's algorithm papers; this is an independent implementation
from the papers' ideas, no libavoid code, which stays rejected on
LGPL grounds). It is deliberately DISTINCT from the patented
interactive-routing methods surveyed in Phase 1 (US 8,542,234;
US 9,082,226; US 12,051,137): plain grid construction over
inflated obstacle borders, no incremental-nudge or
lane-reservation machinery from those claims.
Coordinates are the same absolute space as StructuralGeometry
(top-left boxes, y down).
Orthogonal obstacle-aware edge router (G3L:RTE-011; workstream B4).
Purpose: incremental rerouting during interaction (MR-8): when a node drags, its incident edges must (a) stay anchored at the node BORDER, leaving perpendicular to the attach side, and (b) never route over other nodes. ELK owns from-scratch scene routing; this router owns the interactive path.
Method: the classic sparse orthogonal visibility grid with A* search and a bend penalty, per the published literature on object-avoiding orthogonal connector routing (Wybrow, Marriott & Stuckey's algorithm papers; this is an independent implementation from the papers' ideas, no libavoid code, which stays rejected on LGPL grounds). It is deliberately DISTINCT from the patented interactive-routing methods surveyed in Phase 1 (US 8,542,234; US 9,082,226; US 12,051,137): plain grid construction over inflated obstacle borders, no incremental-nudge or lane-reservation machinery from those claims.
Coordinates are the same absolute space as StructuralGeometry (top-left boxes, y down).