Refactor Graph edge key storage to use JSON serialization

plugins/core/src/graph/duplicate.ts

@@ -105,7 +105,7 @@ export function detectDuplicates(graph: Graph, options: DuplicateOptions = {}) {
 
                     // Ensure n1 < n2 lexicographically to avoid duplicate pairs
                     const [a, b] = n1.url < n2.url ? [n1, n2] : [n2, n1];
-                    const pairKey = `${a.url}|${b.url}`;
+                    const pairKey = Graph.getEdgeKey(a.url, b.url);
 
                     if (checkedPairs.has(pairKey)) continue;
                     checkedPairs.add(pairKey);
@@ -262,7 +262,7 @@ export function detectDuplicates(graph: Graph, options: DuplicateOptions = {}) {
             // Skip self-referential edges caused by repointing
             if (actualSource === actualTarget) continue;
 
-            const edgeKey = `${actualSource}|${actualTarget}`;
+            const edgeKey = Graph.getEdgeKey(actualSource, actualTarget);
             const existingWeight = updatedEdges.get(edgeKey) || 0;
             updatedEdges.set(edgeKey, Math.max(existingWeight, edge.weight)); // deduplicate
         }
@@ -276,7 +276,7 @@ export function detectDuplicates(graph: Graph, options: DuplicateOptions = {}) {
             node.outLinks = 0;
         }
         for (const [edgeKey, _weight] of updatedEdges.entries()) {
-            const [src, tgt] = edgeKey.split('|');
+            const { source: src, target: tgt } = Graph.parseEdgeKey(edgeKey);
             const sn = graph.nodes.get(src);
             const tn = graph.nodes.get(tgt);
             if (sn) sn.outLinks++;

plugins/core/src/graph/graph.ts

@@ -60,7 +60,7 @@ export interface CrawlStats {
 
 export class Graph {
   nodes: Map<string, GraphNode> = new Map();
-  // Using string "source|target" to ensure uniqueness efficiently. Mapping to weight.
+  // Using JSON string of [source, target] to ensure uniqueness. Mapping to weight.
   edges: Map<string, number> = new Map();
   limitReached: boolean = false;
   sessionStats: CrawlStats = {
@@ -73,6 +73,21 @@ export class Graph {
   duplicateClusters: { id: string; type: 'exact' | 'near' | 'template_heavy'; size: number; representative: string; severity: 'low' | 'medium' | 'high' }[] = [];
   contentClusters: ClusterInfo[] = [];
 
+  /**
+   * Generates a unique key for an edge.
+   */
+  static getEdgeKey(source: string, target: string): string {
+    return JSON.stringify([source, target]);
+  }
+
+  /**
+   * Parses an edge key back into source and target.
+   */
+  static parseEdgeKey(key: string): { source: string; target: string } {
+    const [source, target] = JSON.parse(key);
+    return { source, target };
+  }
+
   /**
    * Adds a node to the graph if it doesn't exist.
    * If it exists, updates the status if the new status is non-zero (meaning we crawled it).
@@ -113,7 +128,7 @@ export class Graph {
     const targetNode = this.nodes.get(target);
 
     if (sourceNode && targetNode) {
-      const edgeKey = `${source}|${target}`;
+      const edgeKey = Graph.getEdgeKey(source, target);
       if (!this.edges.has(edgeKey)) {
         this.edges.set(edgeKey, weight);
         sourceNode.outLinks++;
@@ -134,7 +149,7 @@ export class Graph {
 
   getEdges(): GraphEdge[] {
     return Array.from(this.edges.entries()).map(([edge, weight]) => {
-      const [source, target] = edge.split('|');
+      const { source, target } = Graph.parseEdgeKey(edge);
       return { source, target, weight };
     });
   }
@@ -157,7 +172,7 @@ export class Graph {
     }
     if (json.edges) {
       for (const edge of json.edges) {
-        const key = `${edge.source}|${edge.target}`;
+        const key = Graph.getEdgeKey(edge.source, edge.target);
         graph.edges.set(key, edge.weight || 1.0);
       }
     }

plugins/core/tests/duplicate.test.ts

@@ -89,7 +89,7 @@ describe('Duplicate Detection', () => {
         graph.updateNodeData('https://example.com/b', { contentHash: 'h1' });
 
         // Add edge pointing to B
-        graph.edges.set('https://example.com/source|https://example.com/b', 1);
+        graph.addEdge('https://example.com/source', 'https://example.com/b', 1);
 
         // Force A to be the representative by giving it higher inLinks manually, though it's determined dynamically
         graph.nodes.get('https://example.com/a')!.inLinks = 10;
@@ -105,6 +105,6 @@ describe('Duplicate Detection', () => {
         expect(b.collapseInto).toBe('https://example.com/a');
 
         // Check edge transfer
-        expect(graph.edges.has('https://example.com/source|https://example.com/a')).toBe(true);
+        expect(graph.edges.has(Graph.getEdgeKey('https://example.com/source', 'https://example.com/a'))).toBe(true);
     });
 });