Repository navigation
Added
-
Multi-level Dijkstra routing (#413, #414) —
graph.multiLevelIndex(levels:baseRows:baseColumns:)builds a reusable hierarchy of nested grid cells with sparse boundary overlays, andGraph.MultiLevelIndex.shortestPath(from:to:)answers repeated point-to-point queries with a level-aware bidirectional search. Works for directed and undirected graphs and every projection; queries withblockedNodesoredgeFilterfall back to plain Dijkstra.graph.shortestPathViaMultiLevel(from:to:)is a one-shot convenience.if let index = graph.multiLevelIndex(levels: 3, baseRows: 32, baseColumns: 32) { let path = index.shortestPath(from: start, to: end) }
-
New DocC guide: Graph routing algorithms and tuning — explains how the graph, direct searches, contraction, routing indexes, landmark routing and MLD fit together, with guidance on mode-specific graphs, constraints, custom costs and tuning parameters.
-
Real-world routing coverage and benchmarks — MLD is cross-checked against Dijkstra on Munich and Paris fixtures, including cycling-filtered Munich, and the benchmark suite reports index build and query times for Allgäu, Munich (undirected and directed) and Paris.