Native data structures
Each implementation uses the standard-library priority queue idiom its ecosystem expects.
Same algorithm.
Different idioms.
Inspect repository-controlled Dijkstra implementations in six languages. Concept markers link semantic trace events to the lines that produce them.
Read-only by design. Code is never compiled or executed in the browser.
dijkstra.ts31 lines · read only
01export function dijkstra(graph: Graph, start: Id, goal: Id) {02 const distances = new Map<Id, number>([[start, 0]]);03 const parents = new Map<Id, Id>();04 const frontier = new MinHeap<Entry>();05 frontier.push({ nodeId: start, priority: 0 });06 07 while (!frontier.isEmpty()) {08 const current = frontier.pop()!;09 if (current.priority !== distances.get(current.nodeId)) continue;10 if (current.nodeId === goal) break;11 12 for (const edge of graph.outgoing(current.nodeId)) {13 if (edge.closed) continue;14 const candidate = current.priority + edge.weight;15 const previous = distances.get(edge.toId) ?? Infinity;16 17 if (candidate < previous) {18 distances.set(edge.toId, candidate);19 parents.set(edge.toId, current.nodeId);20 frontier.push({ nodeId: edge.toId, priority: candidate });21 }22 }23 }24 25 const path: Id[] = [];26 for (let node: Id | undefined = goal; node; node = parents.get(node)) {27 path.push(node);28 if (node === start) return path.reverse();29 }30 return [];31}Implementation notes
Equivalent does not mean identical.
Stable tie-breaking
Equal-cost candidates use node identifiers so fixtures and trace order remain reproducible.
Shared fixtures
The same graph inputs verify cost, path connectivity, closed roads, and unreachable destinations.