graphwise 1.8.0 → 1.9.0

package/dist/ranking-3ez5m67U.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"ranking-3ez5m67U.js","names":[],"sources":["../src/ranking/parse.ts","../src/ranking/baselines/utils.ts","../src/ranking/baselines/shortest.ts","../src/ranking/baselines/degree-sum.ts","../src/ranking/baselines/widest-path.ts","../src/ranking/baselines/jaccard-arithmetic.ts","../src/ranking/baselines/pagerank.ts","../src/ranking/baselines/betweenness.ts","../src/ranking/baselines/katz.ts","../src/ranking/baselines/communicability.ts","../src/ranking/baselines/resistance-distance.ts","../src/ranking/baselines/random-ranking.ts","../src/ranking/baselines/hitting-time.ts"],"sourcesContent":["/**\n * PARSE (Path-Aware Ranking via Salience Estimation).\n *\n * Ranks discovered paths by computing geometric mean of edge MI scores.\n * Path salience = (∏ MI(uᵢ, uᵢ₊₁))^(1/|path|)\n *\n * This ranking is length-unbiased: shorter paths with strong edges\n * can outrank longer paths with weak edges.\n *\n * @module ranking/parse\n */\n\nimport type { NodeData, EdgeData, ReadableGraph } from \"../graph\";\nimport type { AsyncReadableGraph } from \"../graph/async-interfaces\";\nimport type { ExpansionPath } from \"../expansion/types\";\nimport type { MIFunction, AsyncMIFunction } from \"./mi/types\";\nimport { jaccard } from \"./mi/jaccard\";\nimport { jaccardAsync } from \"./mi/jaccard\";\n\n/**\n * Configuration for PARSE ranking.\n */\nexport interface PARSEConfig<\n\tN extends NodeData = NodeData,\n\tE extends EdgeData = EdgeData,\n> {\n\t/** MI function to use (default: jaccard) */\n\treadonly mi?: MIFunction<N, E>;\n\t/** Minimum epsilon for MI (default: 1e-10) */\n\treadonly epsilon?: number;\n\t/** Whether to include salience scores in result (default: true) */\n\treadonly includeSalience?: boolean;\n}\n\n/**\n * A ranked path with salience score.\n */\nexport interface RankedPath extends ExpansionPath {\n\t/** Salience score (geometric mean of edge MI) */\n\treadonly salience: number;\n}\n\n/**\n * Result of PARSE ranking.\n */\nexport interface PARSEResult {\n\t/** Paths ranked by salience (highest first) */\n\treadonly paths: readonly RankedPath[];\n\t/** Ranking statistics */\n\treadonly stats: {\n\t\t/** Total paths ranked */\n\t\treadonly pathsRanked: number;\n\t\t/** Mean salience */\n\t\treadonly meanSalience: number;\n\t\t/** Median salience */\n\t\treadonly medianSalience: number;\n\t\t/** Maximum salience */\n\t\treadonly maxSalience: number;\n\t\t/** Minimum salience */\n\t\treadonly minSalience: number;\n\t\t/** Ranking duration in milliseconds */\n\t\treadonly durationMs: number;\n\t};\n}\n\n/**\n * Rank paths using PARSE (Path-Aware Ranking via Salience Estimation).\n *\n * Computes geometric mean of edge MI scores for each path,\n * then sorts by salience (highest first).\n *\n * @param graph - Source graph\n * @param paths - Paths to rank\n * @param config - Configuration options\n * @returns Ranked paths with statistics\n */\nexport function parse<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tpaths: readonly ExpansionPath[],\n\tconfig?: PARSEConfig<N, E>,\n): PARSEResult {\n\tconst startTime = performance.now();\n\n\tconst { mi = jaccard, epsilon = 1e-10 } = config ?? {};\n\n\tconst rankedPaths: RankedPath[] = [];\n\n\tfor (const path of paths) {\n\t\tconst salience = computePathSalience(graph, path, mi, epsilon);\n\t\trankedPaths.push({\n\t\t\t...path,\n\t\t\tsalience,\n\t\t});\n\t}\n\n\t// Sort by salience descending\n\trankedPaths.sort((a, b) => b.salience - a.salience);\n\n\tconst endTime = performance.now();\n\n\t// Compute statistics\n\tconst saliences = rankedPaths.map((p) => p.salience);\n\tconst meanSalience =\n\t\tsaliences.length > 0\n\t\t\t? saliences.reduce((a, b) => a + b, 0) / saliences.length\n\t\t\t: 0;\n\tconst sortedSaliences = [...saliences].sort((a, b) => a - b);\n\tconst mid = Math.floor(sortedSaliences.length / 2);\n\tconst medianSalience =\n\t\tsortedSaliences.length > 0\n\t\t\t? sortedSaliences.length % 2 !== 0\n\t\t\t\t? (sortedSaliences[mid] ?? 0)\n\t\t\t\t: ((sortedSaliences[mid - 1] ?? 0) + (sortedSaliences[mid] ?? 0)) / 2\n\t\t\t: 0;\n\tconst maxSalience =\n\t\tsortedSaliences.length > 0\n\t\t\t? (sortedSaliences[sortedSaliences.length - 1] ?? 0)\n\t\t\t: 0;\n\tconst minSalience =\n\t\tsortedSaliences.length > 0 ? (sortedSaliences[0] ?? 0) : 0;\n\n\treturn {\n\t\tpaths: rankedPaths,\n\t\tstats: {\n\t\t\tpathsRanked: rankedPaths.length,\n\t\t\tmeanSalience,\n\t\t\tmedianSalience,\n\t\t\tmaxSalience,\n\t\t\tminSalience,\n\t\t\tdurationMs: endTime - startTime,\n\t\t},\n\t};\n}\n\n/**\n * Configuration for async PARSE ranking.\n */\nexport interface AsyncPARSEConfig<\n\tN extends NodeData = NodeData,\n\tE extends EdgeData = EdgeData,\n> {\n\t/** Async MI function to use (default: jaccardAsync) */\n\treadonly mi?: AsyncMIFunction<N, E>;\n\t/** Minimum epsilon for MI (default: 1e-10) */\n\treadonly epsilon?: number;\n}\n\n/**\n * Rank paths using async PARSE (Path-Aware Ranking via Salience Estimation).\n *\n * Async variant suitable for use with remote or lazy graph data sources.\n * Computes geometric mean of edge MI scores for each path using Promise.all\n * for parallelism, then sorts by salience (highest first).\n *\n * @param graph - Async source graph\n * @param paths - Paths to rank\n * @param config - Configuration options\n * @returns Ranked paths with statistics\n */\nexport async function parseAsync<N extends NodeData, E extends EdgeData>(\n\tgraph: AsyncReadableGraph<N, E>,\n\tpaths: readonly ExpansionPath[],\n\tconfig?: AsyncPARSEConfig<N, E>,\n): Promise<PARSEResult> {\n\tconst startTime = performance.now();\n\n\tconst { mi = jaccardAsync, epsilon = 1e-10 } = config ?? {};\n\n\tconst rankedPaths: RankedPath[] = [];\n\n\tfor (const path of paths) {\n\t\tconst salience = await computePathSalienceAsync(graph, path, mi, epsilon);\n\t\trankedPaths.push({\n\t\t\t...path,\n\t\t\tsalience,\n\t\t});\n\t}\n\n\t// Sort by salience descending\n\trankedPaths.sort((a, b) => b.salience - a.salience);\n\n\tconst endTime = performance.now();\n\n\t// Compute statistics\n\tconst saliences = rankedPaths.map((p) => p.salience);\n\tconst meanSalience =\n\t\tsaliences.length > 0\n\t\t\t? saliences.reduce((a, b) => a + b, 0) / saliences.length\n\t\t\t: 0;\n\tconst sortedSaliences = [...saliences].sort((a, b) => a - b);\n\tconst mid = Math.floor(sortedSaliences.length / 2);\n\tconst medianSalience =\n\t\tsortedSaliences.length > 0\n\t\t\t? sortedSaliences.length % 2 !== 0\n\t\t\t\t? (sortedSaliences[mid] ?? 0)\n\t\t\t\t: ((sortedSaliences[mid - 1] ?? 0) + (sortedSaliences[mid] ?? 0)) / 2\n\t\t\t: 0;\n\tconst maxSalience =\n\t\tsortedSaliences.length > 0\n\t\t\t? (sortedSaliences[sortedSaliences.length - 1] ?? 0)\n\t\t\t: 0;\n\tconst minSalience =\n\t\tsortedSaliences.length > 0 ? (sortedSaliences[0] ?? 0) : 0;\n\n\treturn {\n\t\tpaths: rankedPaths,\n\t\tstats: {\n\t\t\tpathsRanked: rankedPaths.length,\n\t\t\tmeanSalience,\n\t\t\tmedianSalience,\n\t\t\tmaxSalience,\n\t\t\tminSalience,\n\t\t\tdurationMs: endTime - startTime,\n\t\t},\n\t};\n}\n\n/**\n * Compute salience for a single path asynchronously.\n *\n * Uses geometric mean of edge MI scores for length-unbiased ranking.\n * Edge MI values are computed in parallel via Promise.all.\n */\nasync function computePathSalienceAsync<N extends NodeData, E extends EdgeData>(\n\tgraph: AsyncReadableGraph<N, E>,\n\tpath: ExpansionPath,\n\tmi: AsyncMIFunction<N, E>,\n\tepsilon: number,\n): Promise<number> {\n\tconst nodes = path.nodes;\n\n\tif (nodes.length < 2) {\n\t\treturn epsilon;\n\t}\n\n\t// Compute MI for each edge in parallel\n\tconst edgeMIs = await Promise.all(\n\t\tnodes.slice(0, -1).map((source, i) => {\n\t\t\tconst target = nodes[i + 1];\n\t\t\tif (target !== undefined) {\n\t\t\t\treturn mi(graph, source, target);\n\t\t\t}\n\t\t\treturn Promise.resolve(epsilon);\n\t\t}),\n\t);\n\n\tlet productMi = 1;\n\tlet edgeCount = 0;\n\n\tfor (const edgeMi of edgeMIs) {\n\t\tproductMi *= Math.max(epsilon, edgeMi);\n\t\tedgeCount++;\n\t}\n\n\tif (edgeCount === 0) {\n\t\treturn epsilon;\n\t}\n\n\t// Geometric mean\n\tconst salience = Math.pow(productMi, 1 / edgeCount);\n\treturn Math.max(epsilon, Math.min(1, salience));\n}\n\n/**\n * Compute salience for a single path.\n *\n * Uses geometric mean of edge MI scores for length-unbiased ranking.\n */\nfunction computePathSalience<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tpath: ExpansionPath,\n\tmi: MIFunction<N, E>,\n\tepsilon: number,\n): number {\n\tconst nodes = path.nodes;\n\n\tif (nodes.length < 2) {\n\t\treturn epsilon;\n\t}\n\n\t// Compute MI for each edge\n\tlet productMi = 1;\n\tlet edgeCount = 0;\n\n\tfor (let i = 0; i < nodes.length - 1; i++) {\n\t\tconst source = nodes[i];\n\t\tconst target = nodes[i + 1];\n\n\t\tif (source !== undefined && target !== undefined) {\n\t\t\tconst edgeMi = mi(graph, source, target);\n\t\t\tproductMi *= Math.max(epsilon, edgeMi);\n\t\t\tedgeCount++;\n\t\t}\n\t}\n\n\tif (edgeCount === 0) {\n\t\treturn epsilon;\n\t}\n\n\t// Geometric mean\n\tconst salience = Math.pow(productMi, 1 / edgeCount);\n\treturn Math.max(epsilon, Math.min(1, salience));\n}\n","/**\n * Shared utilities for baseline ranking methods.\n *\n * @packageDocumentation\n */\n\nimport type { ExpansionPath } from \"../../expansion/types\";\nimport type { BaselineResult, ScoredPath } from \"./types\";\n\n/**\n * Normalise a set of scored paths and return them sorted highest-first.\n *\n * All scores are normalised relative to the maximum observed score.\n * When `includeScores` is false, raw (un-normalised) scores are preserved.\n * Handles degenerate cases: empty input and all-zero scores.\n *\n * @param paths - Original paths in input order\n * @param scored - Paths paired with their computed scores\n * @param method - Method name to embed in the result\n * @param includeScores - When true, normalise scores to [0, 1]; when false, keep raw scores\n * @returns BaselineResult with ranked paths\n */\nexport function normaliseAndRank(\n\tpaths: readonly ExpansionPath[],\n\tscored: readonly { readonly path: ExpansionPath; readonly score: number }[],\n\tmethod: string,\n\tincludeScores: boolean,\n): BaselineResult {\n\tif (scored.length === 0) {\n\t\treturn { paths: [], method };\n\t}\n\n\tconst maxScore = Math.max(...scored.map((s) => s.score));\n\n\t// Handle zero-max case: all paths get score 0\n\tif (maxScore === 0) {\n\t\treturn {\n\t\t\tpaths: paths.map((path) => ({ ...path, score: 0 })),\n\t\t\tmethod,\n\t\t};\n\t}\n\n\tconst ranked: ScoredPath[] = scored\n\t\t.map(({ path, score }) => ({\n\t\t\t...path,\n\t\t\tscore: includeScores ? score / maxScore : score,\n\t\t}))\n\t\t.sort((a, b) => b.score - a.score);\n\n\treturn { paths: ranked, method };\n}\n","/**\n * Shortest path baseline ranking.\n *\n * Ranks paths by length (shorter = higher score).\n * Score = 1 / length (normalised).\n */\n\nimport type { NodeData, EdgeData, ReadableGraph } from \"../../graph\";\nimport type { ExpansionPath } from \"../../expansion/types\";\nimport type { BaselineConfig, BaselineResult } from \"./types\";\nimport { normaliseAndRank } from \"./utils\";\n\n/**\n * Rank paths by length (shortest first).\n *\n * Score = 1 / path_length, normalised to [0, 1].\n *\n * @param _graph - Source graph (unused for length ranking)\n * @param paths - Paths to rank\n * @param config - Configuration options\n * @returns Ranked paths (shortest first)\n */\nexport function shortest<N extends NodeData, E extends EdgeData>(\n\t_graph: ReadableGraph<N, E>,\n\tpaths: readonly ExpansionPath[],\n\tconfig?: BaselineConfig,\n): BaselineResult {\n\tconst { includeScores = true } = config ?? {};\n\n\tif (paths.length === 0) {\n\t\treturn {\n\t\t\tpaths: [],\n\t\t\tmethod: \"shortest\",\n\t\t};\n\t}\n\n\t// Compute raw scores (1 / length)\n\tconst scored: { path: ExpansionPath; score: number }[] = paths.map(\n\t\t(path) => ({\n\t\t\tpath,\n\t\t\tscore: 1 / path.nodes.length,\n\t\t}),\n\t);\n\n\treturn normaliseAndRank(paths, scored, \"shortest\", includeScores);\n}\n","/**\n * Degree-Sum baseline ranking.\n *\n * Ranks paths by sum of node degrees.\n * Higher degree nodes may indicate more connected (central) nodes.\n * Score = sum(deg(v) for v in path), normalised to [0, 1].\n */\n\nimport type { NodeData, EdgeData, ReadableGraph } from \"../../graph\";\nimport type { ExpansionPath } from \"../../expansion/types\";\nimport type { BaselineConfig, BaselineResult } from \"./types\";\nimport { normaliseAndRank } from \"./utils\";\n\n/**\n * Rank paths by sum of node degrees.\n *\n * @param graph - Source graph\n * @param paths - Paths to rank\n * @param config - Configuration options\n * @returns Ranked paths (highest degree-sum first)\n */\nexport function degreeSum<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tpaths: readonly ExpansionPath[],\n\tconfig?: BaselineConfig,\n): BaselineResult {\n\tconst { includeScores = true } = config ?? {};\n\n\tif (paths.length === 0) {\n\t\treturn {\n\t\t\tpaths: [],\n\t\t\tmethod: \"degree-sum\",\n\t\t};\n\t}\n\n\t// Compute raw scores (sum of degrees)\n\tconst scored: { path: ExpansionPath; score: number }[] = paths.map((path) => {\n\t\tlet degreeSum = 0;\n\t\tfor (const nodeId of path.nodes) {\n\t\t\tdegreeSum += graph.degree(nodeId);\n\t\t}\n\t\treturn { path, score: degreeSum };\n\t});\n\n\treturn normaliseAndRank(paths, scored, \"degree-sum\", includeScores);\n}\n","/**\n * Widest-Path baseline ranking.\n *\n * Ranks paths by bottleneck similarity (minimum edge salience).\n * Uses Jaccard similarity as the edge salience metric.\n * Score = min(jaccard(u, v) for each edge (u,v) in path).\n */\n\nimport type { NodeData, EdgeData, ReadableGraph } from \"../../graph\";\nimport type { ExpansionPath } from \"../../expansion/types\";\nimport type { BaselineConfig, BaselineResult } from \"./types\";\nimport { jaccard } from \"../mi/jaccard\";\nimport { normaliseAndRank } from \"./utils\";\n\n/**\n * Rank paths by widest bottleneck (minimum edge similarity).\n *\n * @param graph - Source graph\n * @param paths - Paths to rank\n * @param config - Configuration options\n * @returns Ranked paths (highest bottleneck first)\n */\nexport function widestPath<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tpaths: readonly ExpansionPath[],\n\tconfig?: BaselineConfig,\n): BaselineResult {\n\tconst { includeScores = true } = config ?? {};\n\n\tif (paths.length === 0) {\n\t\treturn {\n\t\t\tpaths: [],\n\t\t\tmethod: \"widest-path\",\n\t\t};\n\t}\n\n\t// Compute raw scores (minimum edge similarity per path)\n\tconst scored: { path: ExpansionPath; score: number }[] = paths.map((path) => {\n\t\tif (path.nodes.length < 2) {\n\t\t\t// Single-node path: no edges\n\t\t\treturn { path, score: 1 };\n\t\t}\n\n\t\tlet minSimilarity = Number.POSITIVE_INFINITY;\n\t\tfor (let i = 0; i < path.nodes.length - 1; i++) {\n\t\t\tconst source = path.nodes[i];\n\t\t\tconst target = path.nodes[i + 1];\n\t\t\tif (source === undefined || target === undefined) continue;\n\n\t\t\tconst edgeSimilarity = jaccard(graph, source, target);\n\t\t\tminSimilarity = Math.min(minSimilarity, edgeSimilarity);\n\t\t}\n\n\t\t// If no edges were found, default to 1\n\t\tconst score =\n\t\t\tminSimilarity === Number.POSITIVE_INFINITY ? 1 : minSimilarity;\n\t\treturn { path, score };\n\t});\n\n\treturn normaliseAndRank(paths, scored, \"widest-path\", includeScores);\n}\n","/**\n * Jaccard-Arithmetic baseline ranking.\n *\n * Ranks paths by arithmetic mean of edge Jaccard similarities.\n * Contrast with PARSE which uses geometric mean.\n * Score = (1/k) * sum(jaccard(u, v) for each edge (u,v)).\n */\n\nimport type { NodeData, EdgeData, ReadableGraph } from \"../../graph\";\nimport type { ExpansionPath } from \"../../expansion/types\";\nimport type { BaselineConfig, BaselineResult } from \"./types\";\nimport { jaccard } from \"../mi/jaccard\";\nimport { normaliseAndRank } from \"./utils\";\n\n/**\n * Rank paths by arithmetic mean of edge Jaccard similarities.\n *\n * @param graph - Source graph\n * @param paths - Paths to rank\n * @param config - Configuration options\n * @returns Ranked paths (highest arithmetic mean first)\n */\nexport function jaccardArithmetic<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tpaths: readonly ExpansionPath[],\n\tconfig?: BaselineConfig,\n): BaselineResult {\n\tconst { includeScores = true } = config ?? {};\n\n\tif (paths.length === 0) {\n\t\treturn {\n\t\t\tpaths: [],\n\t\t\tmethod: \"jaccard-arithmetic\",\n\t\t};\n\t}\n\n\t// Compute raw scores (arithmetic mean of edge similarities)\n\tconst scored: { path: ExpansionPath; score: number }[] = paths.map((path) => {\n\t\tif (path.nodes.length < 2) {\n\t\t\t// Single-node path: no edges, score = 1\n\t\t\treturn { path, score: 1 };\n\t\t}\n\n\t\tlet similaritySum = 0;\n\t\tlet edgeCount = 0;\n\n\t\tfor (let i = 0; i < path.nodes.length - 1; i++) {\n\t\t\tconst source = path.nodes[i];\n\t\t\tconst target = path.nodes[i + 1];\n\t\t\tif (source === undefined || target === undefined) continue;\n\n\t\t\tconst edgeSimilarity = jaccard(graph, source, target);\n\t\t\tsimilaritySum += edgeSimilarity;\n\t\t\tedgeCount++;\n\t\t}\n\n\t\t// Arithmetic mean\n\t\tconst score = edgeCount > 0 ? similaritySum / edgeCount : 1;\n\t\treturn { path, score };\n\t});\n\n\treturn normaliseAndRank(paths, scored, \"jaccard-arithmetic\", includeScores);\n}\n","/**\n * PageRank baseline ranking.\n *\n * Computes PageRank centrality for all nodes, then sums PR values per path.\n * Uses power iteration: r(v) = (1-d)/N + d * sum(r(u)/deg_out(u) for u->v)\n * Parameters: d=0.85 (damping factor), tolerance=1e-6, max 100 iterations.\n * Score = sum(pagerank(v) for v in path), normalised to [0, 1].\n */\n\nimport type { NodeData, EdgeData, ReadableGraph } from \"../../graph\";\nimport type { ExpansionPath } from \"../../expansion/types\";\nimport type { BaselineConfig, BaselineResult } from \"./types\";\nimport { normaliseAndRank } from \"./utils\";\n\n/**\n * Compute PageRank centrality for all nodes using power iteration.\n *\n * @param graph - Source graph\n * @param damping - Damping factor (default 0.85)\n * @param tolerance - Convergence tolerance (default 1e-6)\n * @param maxIterations - Maximum iterations (default 100)\n * @returns Map of node ID to PageRank value\n */\nfunction computePageRank<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tdamping = 0.85,\n\ttolerance = 1e-6,\n\tmaxIterations = 100,\n): Map<string, number> {\n\tconst nodes = Array.from(graph.nodeIds());\n\tconst n = nodes.length;\n\n\tif (n === 0) {\n\t\treturn new Map();\n\t}\n\n\t// Initialise ranks uniformly\n\tconst ranks = new Map<string, number>();\n\tconst newRanks = new Map<string, number>();\n\tfor (const nodeId of nodes) {\n\t\tranks.set(nodeId, 1 / n);\n\t\tnewRanks.set(nodeId, 0);\n\t}\n\n\t// Power iteration\n\tlet isCurrentRanks = true; // Track which map is current\n\n\tfor (let iteration = 0; iteration < maxIterations; iteration++) {\n\t\tlet maxChange = 0;\n\t\tconst currMap = isCurrentRanks ? ranks : newRanks;\n\t\tconst nextMap = isCurrentRanks ? newRanks : ranks;\n\n\t\tfor (const nodeId of nodes) {\n\t\t\t// Sum contributions from incoming neighbours\n\t\t\tlet incomingSum = 0;\n\n\t\t\tfor (const incomingId of graph.neighbours(nodeId, \"in\")) {\n\t\t\t\tconst incomingRank = currMap.get(incomingId) ?? 0;\n\t\t\t\tconst outDegree = graph.degree(incomingId);\n\t\t\t\tif (outDegree > 0) {\n\t\t\t\t\tincomingSum += incomingRank / outDegree;\n\t\t\t\t}\n\t\t\t}\n\n\t\t\t// PageRank formula\n\t\t\tconst newRank = (1 - damping) / n + damping * incomingSum;\n\t\t\tnextMap.set(nodeId, newRank);\n\n\t\t\t// Track convergence\n\t\t\tconst oldRank = currMap.get(nodeId) ?? 0;\n\t\t\tmaxChange = Math.max(maxChange, Math.abs(newRank - oldRank));\n\t\t}\n\n\t\t// Check convergence before swapping\n\t\tif (maxChange < tolerance) {\n\t\t\tbreak;\n\t\t}\n\n\t\t// Swap buffers and clear the old current map for next iteration\n\t\tisCurrentRanks = !isCurrentRanks;\n\t\tcurrMap.clear();\n\t}\n\n\treturn isCurrentRanks ? ranks : newRanks;\n}\n\n/**\n * Rank paths by sum of PageRank scores.\n *\n * @param graph - Source graph\n * @param paths - Paths to rank\n * @param config - Configuration options\n * @returns Ranked paths (highest PageRank sum first)\n */\nexport function pagerank<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tpaths: readonly ExpansionPath[],\n\tconfig?: BaselineConfig,\n): BaselineResult {\n\tconst { includeScores = true } = config ?? {};\n\n\tif (paths.length === 0) {\n\t\treturn {\n\t\t\tpaths: [],\n\t\t\tmethod: \"pagerank\",\n\t\t};\n\t}\n\n\t// Compute PageRank\n\tconst ranks = computePageRank(graph);\n\n\t// Score paths by sum of node ranks\n\tconst scored: { path: ExpansionPath; score: number }[] = paths.map((path) => {\n\t\tlet prSum = 0;\n\t\tfor (const nodeId of path.nodes) {\n\t\t\tprSum += ranks.get(nodeId) ?? 0;\n\t\t}\n\t\treturn { path, score: prSum };\n\t});\n\n\treturn normaliseAndRank(paths, scored, \"pagerank\", includeScores);\n}\n","/**\n * Betweenness baseline ranking.\n *\n * Computes betweenness centrality using Brandes algorithm O(|V||E|).\n * Score = sum(betweenness(v) for v in path), normalised to [0, 1].\n */\n\nimport type { NodeData, EdgeData, ReadableGraph } from \"../../graph\";\nimport type { ExpansionPath } from \"../../expansion/types\";\nimport type { BaselineConfig, BaselineResult } from \"./types\";\nimport { normaliseAndRank } from \"./utils\";\n\n/**\n * Compute betweenness centrality for all nodes using Brandes algorithm.\n *\n * @param graph - Source graph\n * @returns Map of node ID to betweenness value\n */\nfunction computeBetweenness<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n): Map<string, number> {\n\tconst nodes = Array.from(graph.nodeIds());\n\tconst betweenness = new Map<string, number>();\n\n\t// Initialise all betweenness to 0\n\tfor (const nodeId of nodes) {\n\t\tbetweenness.set(nodeId, 0);\n\t}\n\n\t// For each node as source\n\tfor (const source of nodes) {\n\t\t// BFS to find shortest paths\n\t\tconst predecessors = new Map<string, string[]>();\n\t\tconst distance = new Map<string, number>();\n\t\tconst sigma = new Map<string, number>();\n\t\tconst queue: string[] = [];\n\n\t\t// Initialise\n\t\tfor (const nodeId of nodes) {\n\t\t\tpredecessors.set(nodeId, []);\n\t\t\tdistance.set(nodeId, -1);\n\t\t\tsigma.set(nodeId, 0);\n\t\t}\n\n\t\tdistance.set(source, 0);\n\t\tsigma.set(source, 1);\n\t\tqueue.push(source);\n\n\t\t// BFS\n\t\tfor (const v of queue) {\n\t\t\tconst vDist = distance.get(v) ?? -1;\n\t\t\tconst neighbours = graph.neighbours(v);\n\n\t\t\tfor (const w of neighbours) {\n\t\t\t\tconst wDist = distance.get(w) ?? -1;\n\n\t\t\t\t// First time seeing w?\n\t\t\t\tif (wDist < 0) {\n\t\t\t\t\tdistance.set(w, vDist + 1);\n\t\t\t\t\tqueue.push(w);\n\t\t\t\t}\n\n\t\t\t\t// Shortest path to w through v?\n\t\t\t\tif (wDist === vDist + 1) {\n\t\t\t\t\tconst wSigma = sigma.get(w) ?? 0;\n\t\t\t\t\tconst vSigma = sigma.get(v) ?? 0;\n\t\t\t\t\tsigma.set(w, wSigma + vSigma);\n\n\t\t\t\t\tconst wPred = predecessors.get(w) ?? [];\n\t\t\t\t\twPred.push(v);\n\t\t\t\t\tpredecessors.set(w, wPred);\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\n\t\t// Accumulation\n\t\tconst delta = new Map<string, number>();\n\t\tfor (const nodeId of nodes) {\n\t\t\tdelta.set(nodeId, 0);\n\t\t}\n\n\t\t// Process in reverse order of distance\n\t\tconst sorted = [...nodes].sort((a, b) => {\n\t\t\tconst aD = distance.get(a) ?? -1;\n\t\t\tconst bD = distance.get(b) ?? -1;\n\t\t\treturn bD - aD;\n\t\t});\n\n\t\tfor (const w of sorted) {\n\t\t\tif (w === source) continue;\n\n\t\t\tconst wDelta = delta.get(w) ?? 0;\n\t\t\tconst wSigma = sigma.get(w) ?? 0;\n\n\t\t\tconst wPred = predecessors.get(w) ?? [];\n\t\t\tfor (const v of wPred) {\n\t\t\t\tconst vSigma = sigma.get(v) ?? 0;\n\t\t\t\tconst vDelta = delta.get(v) ?? 0;\n\n\t\t\t\tif (wSigma > 0) {\n\t\t\t\t\tdelta.set(v, vDelta + (vSigma / wSigma) * (1 + wDelta));\n\t\t\t\t}\n\t\t\t}\n\n\t\t\tif (w !== source) {\n\t\t\t\tconst current = betweenness.get(w) ?? 0;\n\t\t\t\tbetweenness.set(w, current + wDelta);\n\t\t\t}\n\t\t}\n\t}\n\n\treturn betweenness;\n}\n\n/**\n * Rank paths by sum of betweenness scores.\n *\n * @param graph - Source graph\n * @param paths - Paths to rank\n * @param config - Configuration options\n * @returns Ranked paths (highest betweenness sum first)\n */\nexport function betweenness<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tpaths: readonly ExpansionPath[],\n\tconfig?: BaselineConfig,\n): BaselineResult {\n\tconst { includeScores = true } = config ?? {};\n\n\tif (paths.length === 0) {\n\t\treturn {\n\t\t\tpaths: [],\n\t\t\tmethod: \"betweenness\",\n\t\t};\n\t}\n\n\t// Compute betweenness\n\tconst bcMap = computeBetweenness(graph);\n\n\t// Score paths by sum of betweenness\n\tconst scored: { path: ExpansionPath; score: number }[] = paths.map((path) => {\n\t\tlet bcSum = 0;\n\t\tfor (const nodeId of path.nodes) {\n\t\t\tbcSum += bcMap.get(nodeId) ?? 0;\n\t\t}\n\t\treturn { path, score: bcSum };\n\t});\n\n\treturn normaliseAndRank(paths, scored, \"betweenness\", includeScores);\n}\n","/**\n * Katz baseline ranking.\n *\n * Truncated Katz centrality: score(s,t) = sum_{k=1}^{K} beta^k * walks_k(s,t)\n * Parameters: K=5 (truncation depth), beta=0.005 (safe damping < 1/lambda_1).\n * For path scoring: score(P) = katz(P.start, P.end), normalised to [0, 1].\n */\n\nimport type { NodeData, EdgeData, ReadableGraph } from \"../../graph\";\nimport type { ExpansionPath } from \"../../expansion/types\";\nimport type { BaselineConfig, BaselineResult } from \"./types\";\nimport { normaliseAndRank } from \"./utils\";\n\n/**\n * Compute truncated Katz centrality between two nodes.\n *\n * Uses iterative matrix-vector products to avoid full matrix powers.\n * score(s,t) = sum_{k=1}^{K} beta^k * walks_k(s,t)\n *\n * @param graph - Source graph\n * @param source - Source node ID\n * @param target - Target node ID\n * @param k - Truncation depth (default 5)\n * @param beta - Attenuation factor (default 0.005)\n * @returns Katz score\n */\nfunction computeKatz<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tsource: string,\n\ttarget: string,\n\tk = 5,\n\tbeta = 0.005,\n): number {\n\tconst nodes = Array.from(graph.nodeIds());\n\tconst nodeToIdx = new Map<string, number>();\n\tnodes.forEach((nodeId, idx) => {\n\t\tnodeToIdx.set(nodeId, idx);\n\t});\n\n\tconst n = nodes.length;\n\tif (n === 0) {\n\t\treturn 0;\n\t}\n\n\tconst sourceIdx = nodeToIdx.get(source);\n\tconst targetIdx = nodeToIdx.get(target);\n\n\tif (sourceIdx === undefined || targetIdx === undefined) {\n\t\treturn 0;\n\t}\n\n\t// Current column of A^depth (number of walks of length depth from each node to target)\n\tlet walks = new Float64Array(n);\n\twalks[targetIdx] = 1; // Base case: walks[target] = 1\n\n\tlet katzScore = 0;\n\n\t// Iterate from depth 1 to k\n\tfor (let depth = 1; depth <= k; depth++) {\n\t\t// Multiply by adjacency matrix: walks_next[i] = sum_j A[i,j] * walks[j]\n\t\tconst walksNext = new Float64Array(n);\n\n\t\tfor (const sourceNode of nodes) {\n\t\t\tconst srcIdx = nodeToIdx.get(sourceNode);\n\t\t\tif (srcIdx === undefined) continue;\n\n\t\t\tconst neighbours = graph.neighbours(sourceNode);\n\t\t\tfor (const neighbourId of neighbours) {\n\t\t\t\tconst nIdx = nodeToIdx.get(neighbourId);\n\t\t\t\tif (nIdx === undefined) continue;\n\n\t\t\t\twalksNext[srcIdx] = (walksNext[srcIdx] ?? 0) + (walks[nIdx] ?? 0);\n\t\t\t}\n\t\t}\n\n\t\t// Add contribution: beta^depth * walks_depth[source]\n\t\tconst walkCount = walksNext[sourceIdx] ?? 0;\n\t\tkatzScore += Math.pow(beta, depth) * walkCount;\n\n\t\twalks = walksNext;\n\t}\n\n\treturn katzScore;\n}\n\n/**\n * Rank paths by Katz centrality between endpoints.\n *\n * @param graph - Source graph\n * @param paths - Paths to rank\n * @param config - Configuration options\n * @returns Ranked paths (highest Katz score first)\n */\nexport function katz<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tpaths: readonly ExpansionPath[],\n\tconfig?: BaselineConfig,\n): BaselineResult {\n\tconst { includeScores = true } = config ?? {};\n\n\tif (paths.length === 0) {\n\t\treturn {\n\t\t\tpaths: [],\n\t\t\tmethod: \"katz\",\n\t\t};\n\t}\n\n\t// Score paths by Katz between endpoints\n\tconst scored: { path: ExpansionPath; score: number }[] = paths.map((path) => {\n\t\tconst source = path.nodes[0];\n\t\tconst target = path.nodes[path.nodes.length - 1];\n\n\t\tif (source === undefined || target === undefined) {\n\t\t\treturn { path, score: 0 };\n\t\t}\n\n\t\tconst katzScore = computeKatz(graph, source, target);\n\t\treturn { path, score: katzScore };\n\t});\n\n\treturn normaliseAndRank(paths, scored, \"katz\", includeScores);\n}\n","/**\n * Communicability baseline ranking.\n *\n * Computes communicability between nodes using truncated Taylor series.\n * (e^A)_{s,t} ≈ sum_{k=0}^{15} A^k_{s,t} / k!\n * For path scoring: score(P) = communicability(P.start, P.end), normalised to [0, 1].\n */\n\nimport type { NodeData, EdgeData, ReadableGraph } from \"../../graph\";\nimport type { ExpansionPath } from \"../../expansion/types\";\nimport type { BaselineConfig, BaselineResult } from \"./types\";\nimport { normaliseAndRank } from \"./utils\";\n\n/**\n * Compute truncated communicability between two nodes.\n *\n * Uses Taylor series expansion: (e^A)_{s,t} ≈ sum_{k=0}^{K} A^k_{s,t} / k!\n *\n * @param graph - Source graph\n * @param source - Source node ID\n * @param target - Target node ID\n * @param k - Truncation depth (default 15)\n * @returns Communicability score\n */\nfunction computeCommunicability<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tsource: string,\n\ttarget: string,\n\tk = 15,\n): number {\n\tconst nodes = Array.from(graph.nodeIds());\n\tconst nodeToIdx = new Map<string, number>();\n\tnodes.forEach((nodeId, idx) => {\n\t\tnodeToIdx.set(nodeId, idx);\n\t});\n\n\tconst n = nodes.length;\n\tif (n === 0) {\n\t\treturn 0;\n\t}\n\n\tconst sourceIdx = nodeToIdx.get(source);\n\tconst targetIdx = nodeToIdx.get(target);\n\n\tif (sourceIdx === undefined || targetIdx === undefined) {\n\t\treturn 0;\n\t}\n\n\t// Current column of A^depth\n\tlet walks = new Float64Array(n);\n\twalks[targetIdx] = 1; // Base case: walks[target] = 1\n\n\t// Compute sum: sum_{k=0}^{K} A^k_{s,t} / k!\n\tlet commScore = walks[sourceIdx] ?? 0; // k=0 term (identity matrix): A^0 = I, so [I]_{s,t} = delta_{s,t}\n\n\tlet factorial = 1;\n\n\t// Iterate from k=1 to k\n\tfor (let depth = 1; depth <= k; depth++) {\n\t\t// Multiply by adjacency matrix: walks_next[i] = sum_j A[i,j] * walks[j]\n\t\tconst walksNext = new Float64Array(n);\n\n\t\tfor (const fromNode of nodes) {\n\t\t\tconst fromIdx = nodeToIdx.get(fromNode);\n\t\t\tif (fromIdx === undefined) continue;\n\n\t\t\tconst neighbours = graph.neighbours(fromNode);\n\t\t\tfor (const toNodeId of neighbours) {\n\t\t\t\tconst toIdx = nodeToIdx.get(toNodeId);\n\t\t\t\tif (toIdx === undefined) continue;\n\n\t\t\t\twalksNext[fromIdx] = (walksNext[fromIdx] ?? 0) + (walks[toIdx] ?? 0);\n\t\t\t}\n\t\t}\n\n\t\tfactorial *= depth;\n\n\t\t// Add contribution: A^depth[s,t] / k!\n\t\tcommScore += (walksNext[sourceIdx] ?? 0) / factorial;\n\n\t\twalks = walksNext;\n\t}\n\n\treturn commScore;\n}\n\n/**\n * Rank paths by communicability between endpoints.\n *\n * @param graph - Source graph\n * @param paths - Paths to rank\n * @param config - Configuration options\n * @returns Ranked paths (highest communicability first)\n */\nexport function communicability<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tpaths: readonly ExpansionPath[],\n\tconfig?: BaselineConfig,\n): BaselineResult {\n\tconst { includeScores = true } = config ?? {};\n\n\tif (paths.length === 0) {\n\t\treturn {\n\t\t\tpaths: [],\n\t\t\tmethod: \"communicability\",\n\t\t};\n\t}\n\n\t// Score paths by communicability between endpoints\n\tconst scored: { path: ExpansionPath; score: number }[] = paths.map((path) => {\n\t\tconst source = path.nodes[0];\n\t\tconst target = path.nodes[path.nodes.length - 1];\n\n\t\tif (source === undefined || target === undefined) {\n\t\t\treturn { path, score: 0 };\n\t\t}\n\n\t\tconst commScore = computeCommunicability(graph, source, target);\n\t\treturn { path, score: commScore };\n\t});\n\n\treturn normaliseAndRank(paths, scored, \"communicability\", includeScores);\n}\n","/**\n * Resistance-Distance baseline ranking.\n *\n * Computes effective resistance via Laplacian pseudoinverse (dense, small graphs only).\n * For path scoring: score(P) = 1 / resistance(P.start, P.end), normalised to [0, 1].\n * Size guard: throws if nodeCount > 5000 (O(n^3) complexity).\n */\n\nimport type { NodeData, EdgeData, ReadableGraph } from \"../../graph\";\nimport type { ExpansionPath } from \"../../expansion/types\";\nimport type { BaselineConfig, BaselineResult } from \"./types\";\nimport { normaliseAndRank } from \"./utils\";\n\n/**\n * Compute effective resistance between two nodes via Laplacian pseudoinverse.\n *\n * Resistance = L^+_{s,s} + L^+_{t,t} - 2*L^+_{s,t}\n * where L^+ is the pseudoinverse of the Laplacian matrix.\n *\n * @param graph - Source graph\n * @param source - Source node ID\n * @param target - Target node ID\n * @returns Effective resistance\n */\nfunction computeResistance<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tsource: string,\n\ttarget: string,\n): number {\n\tconst nodes = Array.from(graph.nodeIds());\n\tconst nodeToIdx = new Map<string, number>();\n\tnodes.forEach((nodeId, idx) => {\n\t\tnodeToIdx.set(nodeId, idx);\n\t});\n\n\tconst n = nodes.length;\n\n\tif (n === 0 || n > 5000) {\n\t\tthrow new Error(\n\t\t\t`Cannot compute resistance distance: graph too large (${String(n)} nodes). Maximum 5000.`,\n\t\t);\n\t}\n\n\tconst sourceIdx = nodeToIdx.get(source);\n\tconst targetIdx = nodeToIdx.get(target);\n\n\tif (sourceIdx === undefined || targetIdx === undefined) {\n\t\treturn 0;\n\t}\n\n\t// Build Laplacian matrix: L = D - A\n\tconst L: number[][] = Array.from({ length: n }, () =>\n\t\tArray.from({ length: n }, () => 0),\n\t);\n\n\tfor (let i = 0; i < n; i++) {\n\t\tconst nodeId = nodes[i];\n\t\tif (nodeId === undefined) continue;\n\n\t\tconst degree = graph.degree(nodeId);\n\t\tconst row = L[i];\n\t\tif (row !== undefined) {\n\t\t\trow[i] = degree; // Diagonal\n\t\t}\n\n\t\tconst neighbours = graph.neighbours(nodeId);\n\t\tfor (const neighbourId of neighbours) {\n\t\t\tconst j = nodeToIdx.get(neighbourId);\n\t\t\tif (j !== undefined && row !== undefined) {\n\t\t\t\trow[j] = -1; // Off-diagonal\n\t\t\t}\n\t\t}\n\t}\n\n\t// Compute pseudoinverse using Moore-Penrose approach (simplified)\n\t// For small graphs, use LU decomposition with diagonal adjustment\n\tconst Lpinv = pinv(L);\n\n\t// Resistance = L^+_{s,s} + L^+_{t,t} - 2*L^+_{s,t}\n\tconst resistance =\n\t\t(Lpinv[sourceIdx]?.[sourceIdx] ?? 0) +\n\t\t(Lpinv[targetIdx]?.[targetIdx] ?? 0) -\n\t\t2 * (Lpinv[sourceIdx]?.[targetIdx] ?? 0);\n\n\t// Clamp to positive (numerical stability)\n\treturn Math.max(resistance, 1e-10);\n}\n\n/**\n * Compute Moore-Penrose pseudoinverse of a matrix.\n * Simplified implementation for small dense matrices.\n *\n * @param A - Square matrix\n * @returns Pseudoinverse A^+\n */\nfunction pinv(A: number[][]): number[][] {\n\tconst n = A.length;\n\tif (n === 0) return [];\n\n\t// Create copy for singular value computation\n\tconst M = A.map((row) => [...row]);\n\n\t// Simplified: add small regularisation to diagonal before inversion\n\tconst epsilon = 1e-10;\n\tfor (let i = 0; i < n; i++) {\n\t\tconst row = M[i];\n\t\tif (row !== undefined) {\n\t\t\trow[i] = (row[i] ?? 0) + epsilon;\n\t\t}\n\t}\n\n\t// Gaussian elimination with partial pivoting to compute inverse\n\tconst Minv = gaussianInverse(M);\n\n\treturn Minv;\n}\n\n/**\n * Compute matrix inverse using Gaussian elimination with partial pivoting.\n *\n * @param A - Matrix to invert\n * @returns Inverted matrix\n */\nfunction gaussianInverse(A: number[][]): number[][] {\n\tconst n = A.length;\n\n\t// Create augmented matrix [A | I]\n\tconst aug: number[][] = A.map((row, i) => {\n\t\tconst identity: number[] = Array.from({ length: n }, (_, j) =>\n\t\t\ti === j ? 1 : 0,\n\t\t);\n\t\tconst combined: number[] = [...row, ...identity];\n\t\treturn combined;\n\t});\n\n\t// Forward elimination with partial pivoting\n\tfor (let col = 0; col < n; col++) {\n\t\t// Find pivot\n\t\tlet maxRow = col;\n\t\tfor (let row = col + 1; row < n; row++) {\n\t\t\tconst currentRow = aug[row];\n\t\t\tconst maxRowRef = aug[maxRow];\n\t\t\tif (\n\t\t\t\tcurrentRow !== undefined &&\n\t\t\t\tmaxRowRef !== undefined &&\n\t\t\t\tMath.abs(currentRow[col] ?? 0) > Math.abs(maxRowRef[col] ?? 0)\n\t\t\t) {\n\t\t\t\tmaxRow = row;\n\t\t\t}\n\t\t}\n\n\t\t// Swap rows\n\t\tconst currentCol = aug[col];\n\t\tconst maxRowAug = aug[maxRow];\n\t\tif (currentCol !== undefined && maxRowAug !== undefined) {\n\t\t\taug[col] = maxRowAug;\n\t\t\taug[maxRow] = currentCol;\n\t\t}\n\n\t\t// Scale pivot row\n\t\tconst pivotRow = aug[col];\n\t\tconst pivot = pivotRow?.[col];\n\t\tif (pivot === undefined || Math.abs(pivot) < 1e-12) {\n\t\t\tcontinue; // Skip singular column\n\t\t}\n\n\t\tif (pivotRow !== undefined) {\n\t\t\tfor (let j = col; j < 2 * n; j++) {\n\t\t\t\tpivotRow[j] = (pivotRow[j] ?? 0) / pivot;\n\t\t\t}\n\t\t}\n\n\t\t// Eliminate below and above\n\t\tfor (let row = 0; row < n; row++) {\n\t\t\tif (row === col) continue;\n\n\t\t\tconst eliminationRow = aug[row];\n\t\t\tconst factor = eliminationRow?.[col] ?? 0;\n\t\t\tif (eliminationRow !== undefined && pivotRow !== undefined) {\n\t\t\t\tfor (let j = col; j < 2 * n; j++) {\n\t\t\t\t\teliminationRow[j] =\n\t\t\t\t\t\t(eliminationRow[j] ?? 0) - factor * (pivotRow[j] ?? 0);\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\t}\n\n\t// Extract inverse (right half of augmented matrix)\n\tconst Ainv: number[][] = [];\n\tfor (let i = 0; i < n; i++) {\n\t\tconst row = aug[i];\n\t\tAinv[i] = (row?.slice(n) ?? []).map((v) => v);\n\t}\n\n\treturn Ainv;\n}\n\n/**\n * Rank paths by reciprocal of resistance distance between endpoints.\n *\n * @param graph - Source graph\n * @param paths - Paths to rank\n * @param config - Configuration options\n * @returns Ranked paths (highest conductance first)\n */\nexport function resistanceDistance<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tpaths: readonly ExpansionPath[],\n\tconfig?: BaselineConfig,\n): BaselineResult {\n\tconst { includeScores = true } = config ?? {};\n\n\tif (paths.length === 0) {\n\t\treturn {\n\t\t\tpaths: [],\n\t\t\tmethod: \"resistance-distance\",\n\t\t};\n\t}\n\n\t// Check graph size\n\tconst nodeCount = Array.from(graph.nodeIds()).length;\n\tif (nodeCount > 5000) {\n\t\tthrow new Error(\n\t\t\t`Cannot rank paths: graph too large (${String(nodeCount)} nodes). Resistance distance requires O(n^3) computation; maximum 5000 nodes.`,\n\t\t);\n\t}\n\n\t// Score paths by conductance (1 / resistance)\n\tconst scored: { path: ExpansionPath; score: number }[] = paths.map((path) => {\n\t\tconst source = path.nodes[0];\n\t\tconst target = path.nodes[path.nodes.length - 1];\n\n\t\tif (source === undefined || target === undefined) {\n\t\t\treturn { path, score: 0 };\n\t\t}\n\n\t\tconst resistance = computeResistance(graph, source, target);\n\t\t// Score = conductance = 1 / resistance\n\t\tconst score = 1 / resistance;\n\t\treturn { path, score };\n\t});\n\n\treturn normaliseAndRank(paths, scored, \"resistance-distance\", includeScores);\n}\n","/**\n * Random-Ranking baseline.\n *\n * Null hypothesis baseline: ranks paths by seeded random scores.\n * Uses deterministic hash for reproducibility.\n * Score = seededRandom(nodes.join(','), seed), normalised to [0, 1].\n */\n\nimport type { NodeData, EdgeData, ReadableGraph } from \"../../graph\";\nimport type { ExpansionPath } from \"../../expansion/types\";\nimport type { BaselineConfig, BaselineResult } from \"./types\";\nimport { normaliseAndRank } from \"./utils\";\n\n/**\n * Configuration for random ranking.\n */\ninterface RandomRankingConfig extends BaselineConfig {\n\t/** Random seed for deterministic reproducibility */\n\treadonly seed?: number;\n}\n\n/**\n * Deterministic seeded random number generator.\n * Uses FNV-1a-like hash for input → [0, 1] output.\n *\n * @param input - String to hash\n * @param seed - Random seed for reproducibility\n * @returns Deterministic random value in [0, 1]\n */\nfunction seededRandom(input: string, seed = 0): number {\n\tlet h = seed;\n\tfor (let i = 0; i < input.length; i++) {\n\t\th = Math.imul(h ^ input.charCodeAt(i), 0x9e3779b9);\n\n\t\th ^= h >>> 16;\n\t}\n\n\treturn (h >>> 0) / 0xffffffff;\n}\n\n/**\n * Rank paths randomly (null hypothesis baseline).\n *\n * @param _graph - Source graph (unused)\n * @param paths - Paths to rank\n * @param config - Configuration options\n * @returns Ranked paths (randomly ordered)\n */\nexport function randomRanking<N extends NodeData, E extends EdgeData>(\n\t_graph: ReadableGraph<N, E>,\n\tpaths: readonly ExpansionPath[],\n\tconfig?: RandomRankingConfig,\n): BaselineResult {\n\tconst { includeScores = true, seed = 0 } = config ?? {};\n\n\tif (paths.length === 0) {\n\t\treturn {\n\t\t\tpaths: [],\n\t\t\tmethod: \"random\",\n\t\t};\n\t}\n\n\t// Score paths by seeded random hash of node list\n\tconst scored: { path: ExpansionPath; score: number }[] = paths.map((path) => {\n\t\tconst nodesKey = path.nodes.join(\",\");\n\t\tconst score = seededRandom(nodesKey, seed);\n\t\treturn { path, score };\n\t});\n\n\treturn normaliseAndRank(paths, scored, \"random\", includeScores);\n}\n","/**\n * Hitting-time ranking baseline.\n *\n * Ranks paths by the inverse of the expected number of steps\n * in a random walk from source to target.\n *\n * Score = 1 / hittingTime(source, target)\n *\n * Two computation modes:\n * - Approximate: Monte Carlo random walk simulation (default, efficient for large graphs)\n * - Exact: Fundamental matrix approach (for small graphs)\n * - Auto: Automatic mode selection (switches at ~100 nodes)\n */\n\nimport type { NodeData, EdgeData, ReadableGraph } from \"../../graph\";\nimport type { ExpansionPath } from \"../../expansion/types\";\nimport type { BaselineConfig, BaselineResult } from \"./types\";\nimport { normaliseAndRank } from \"./utils\";\n\n/**\n * Configuration for hitting-time ranking.\n */\ninterface HittingTimeConfig extends BaselineConfig {\n\t/** Computation mode: \"exact\", \"approximate\", or \"auto\" (default: \"auto\") */\n\treadonly mode?: \"exact\" | \"approximate\" | \"auto\";\n\t/** Number of Monte Carlo walks for approximate mode (default: 1000) */\n\treadonly walks?: number;\n\t/** Maximum steps per walk (default: 10000) */\n\treadonly maxSteps?: number;\n\t/** Random seed for reproducibility (default: 42) */\n\treadonly seed?: number;\n}\n\n/**\n * Seeded deterministic random number generator (LCG).\n * Suitable for reproducible random walk simulation.\n */\nclass SeededRNG {\n\tprivate state: number;\n\n\tconstructor(seed: number) {\n\t\tthis.state = seed;\n\t}\n\n\t/**\n\t * Generate next pseudorandom value in [0, 1).\n\t */\n\tnext(): number {\n\t\t// Linear congruential generator: standard MINSTD parameters\n\t\tthis.state = (this.state * 1103515245 + 12345) & 0x7fffffff;\n\t\treturn this.state / 0x7fffffff;\n\t}\n}\n\n/**\n * Compute hitting time via Monte Carlo random walk simulation.\n *\n * @param graph - Source graph\n * @param source - Source node ID\n * @param target - Target node ID\n * @param walks - Number of walks to simulate\n * @param maxSteps - Maximum steps per walk\n * @param rng - Seeded RNG instance\n * @returns Average hitting time across walks\n */\nfunction computeHittingTimeApproximate<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tsource: string,\n\ttarget: string,\n\twalks: number,\n\tmaxSteps: number,\n\trng: SeededRNG,\n): number {\n\tif (source === target) {\n\t\treturn 0; // Hitting time from a node to itself is 0\n\t}\n\n\tlet totalSteps = 0;\n\tlet successfulWalks = 0;\n\n\tfor (let w = 0; w < walks; w++) {\n\t\tlet current: string = source;\n\t\tlet steps = 0;\n\n\t\twhile (current !== target && steps < maxSteps) {\n\t\t\tconst neighbours = Array.from(graph.neighbours(current));\n\t\t\tif (neighbours.length === 0) {\n\t\t\t\t// Stuck in sink node\n\t\t\t\tbreak;\n\t\t\t}\n\n\t\t\t// Uniformly choose next neighbour\n\t\t\tconst nextIdx = Math.floor(rng.next() * neighbours.length);\n\t\t\tconst nextNode = neighbours[nextIdx];\n\t\t\tif (nextNode === undefined) {\n\t\t\t\tbreak;\n\t\t\t}\n\t\t\tcurrent = nextNode;\n\t\t\tsteps++;\n\t\t}\n\n\t\tif (current === target) {\n\t\t\ttotalSteps += steps;\n\t\t\tsuccessfulWalks++;\n\t\t}\n\t}\n\n\t// Return average if any walks succeeded\n\tif (successfulWalks > 0) {\n\t\treturn totalSteps / successfulWalks;\n\t}\n\n\t// Return maxSteps as estimate if no walks succeeded\n\treturn maxSteps;\n}\n\n/**\n * Compute hitting time via exact fundamental matrix method.\n *\n * For small graphs, computes exact expected hitting times using\n * the fundamental matrix of the random walk.\n *\n * @param graph - Source graph\n * @param source - Source node ID\n * @param target - Target node ID\n * @returns Exact hitting time (or approximation if convergence fails)\n */\nfunction computeHittingTimeExact<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tsource: string,\n\ttarget: string,\n): number {\n\tif (source === target) {\n\t\treturn 0;\n\t}\n\n\tconst nodes = Array.from(graph.nodeIds());\n\tconst nodeToIdx = new Map<string, number>();\n\tnodes.forEach((nodeId, idx) => {\n\t\tnodeToIdx.set(nodeId, idx);\n\t});\n\n\tconst n = nodes.length;\n\tconst sourceIdx = nodeToIdx.get(source);\n\tconst targetIdx = nodeToIdx.get(target);\n\n\tif (sourceIdx === undefined || targetIdx === undefined) {\n\t\treturn 0;\n\t}\n\n\t// Build transition matrix P with absorbing target\n\tconst P: number[][] = [];\n\tfor (let i = 0; i < n; i++) {\n\t\tconst row: number[] = [];\n\t\tfor (let j = 0; j < n; j++) {\n\t\t\trow[j] = 0;\n\t\t}\n\t\tP[i] = row;\n\t}\n\n\tfor (const nodeId of nodes) {\n\t\tconst idx = nodeToIdx.get(nodeId);\n\t\tif (idx === undefined) continue;\n\n\t\tconst pRow = P[idx];\n\t\tif (pRow === undefined) continue;\n\n\t\tif (idx === targetIdx) {\n\t\t\t// Target is absorbing state\n\t\t\tpRow[idx] = 1;\n\t\t} else {\n\t\t\tconst neighbours = Array.from(graph.neighbours(nodeId));\n\t\t\tconst degree = neighbours.length;\n\n\t\t\tif (degree > 0) {\n\t\t\t\tfor (const neighbourId of neighbours) {\n\t\t\t\t\tconst nIdx = nodeToIdx.get(neighbourId);\n\t\t\t\t\tif (nIdx !== undefined) {\n\t\t\t\t\t\tpRow[nIdx] = 1 / degree;\n\t\t\t\t\t}\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\t}\n\n\t// Compute fundamental matrix N = (I - Q)^(-1)\n\t// where Q is the submatrix of P excluding the absorbing state\n\tconst transientIndices: number[] = [];\n\n\tfor (let i = 0; i < n; i++) {\n\t\tif (i !== targetIdx) {\n\t\t\ttransientIndices.push(i);\n\t\t}\n\t}\n\n\tconst m = transientIndices.length;\n\tconst Q: number[][] = [];\n\tfor (let i = 0; i < m; i++) {\n\t\tconst row: number[] = [];\n\t\tfor (let j = 0; j < m; j++) {\n\t\t\trow[j] = 0;\n\t\t}\n\t\tQ[i] = row;\n\t}\n\n\tfor (let i = 0; i < m; i++) {\n\t\tconst qRow = Q[i];\n\t\tif (qRow === undefined) continue;\n\t\tconst origI = transientIndices[i];\n\t\tif (origI === undefined) continue;\n\t\tconst pRow = P[origI];\n\t\tif (pRow === undefined) continue;\n\n\t\tfor (let j = 0; j < m; j++) {\n\t\t\tconst origJ = transientIndices[j];\n\t\t\tif (origJ === undefined) continue;\n\t\t\tqRow[j] = pRow[origJ] ?? 0;\n\t\t}\n\t}\n\n\t// Compute I - Q\n\tconst IMQ: number[][] = [];\n\tfor (let i = 0; i < m; i++) {\n\t\tconst row: number[] = [];\n\t\tfor (let j = 0; j < m; j++) {\n\t\t\trow[j] = i === j ? 1 : 0;\n\t\t}\n\t\tIMQ[i] = row;\n\t}\n\n\tfor (let i = 0; i < m; i++) {\n\t\tconst imqRow = IMQ[i];\n\t\tif (imqRow === undefined) continue;\n\t\tconst qRow = Q[i];\n\n\t\tfor (let j = 0; j < m; j++) {\n\t\t\tconst qVal = qRow?.[j] ?? 0;\n\t\t\timqRow[j] = (i === j ? 1 : 0) - qVal;\n\t\t}\n\t}\n\n\t// Invert (I - Q) using Gaussian elimination\n\tconst N = invertMatrix(IMQ);\n\n\tif (N === null) {\n\t\t// Fallback if inversion fails\n\t\treturn 1;\n\t}\n\n\t// Hitting time from source to target = sum of row corresponding to source\n\tconst sourceTransientIdx = transientIndices.indexOf(sourceIdx);\n\tif (sourceTransientIdx < 0) {\n\t\treturn 0; // Source is already the target\n\t}\n\n\tlet hittingTime = 0;\n\tconst row = N[sourceTransientIdx];\n\tif (row !== undefined) {\n\t\tfor (const val of row) {\n\t\t\thittingTime += val;\n\t\t}\n\t}\n\n\treturn hittingTime;\n}\n\n/**\n * Invert a square matrix using Gaussian elimination with partial pivoting.\n *\n * @param matrix - Input matrix (n × n)\n * @returns Inverted matrix, or null if singular\n */\nfunction invertMatrix(matrix: number[][]): number[][] | null {\n\tconst n = matrix.length;\n\tconst aug: number[][] = [];\n\n\t// Create augmented matrix [A | I]\n\tfor (let i = 0; i < n; i++) {\n\t\tconst row: number[] = [];\n\t\tconst matRow = matrix[i];\n\n\t\tfor (let j = 0; j < n; j++) {\n\t\t\trow[j] = matRow?.[j] ?? 0;\n\t\t}\n\t\tfor (let j = 0; j < n; j++) {\n\t\t\trow[n + j] = i === j ? 1 : 0;\n\t\t}\n\t\taug[i] = row;\n\t}\n\n\t// Forward elimination with partial pivoting\n\tfor (let col = 0; col < n; col++) {\n\t\t// Find pivot\n\t\tlet pivotRow = col;\n\t\tconst pivotCol = aug[pivotRow];\n\t\tif (pivotCol === undefined) return null;\n\n\t\tfor (let row = col + 1; row < n; row++) {\n\t\t\tconst currRowVal = aug[row]?.[col] ?? 0;\n\t\t\tconst pivotRowVal = pivotCol[col] ?? 0;\n\t\t\tif (Math.abs(currRowVal) > Math.abs(pivotRowVal)) {\n\t\t\t\tpivotRow = row;\n\t\t\t}\n\t\t}\n\n\t\t// Check for singular matrix\n\t\tconst augPivot = aug[pivotRow];\n\t\tif (augPivot === undefined || Math.abs(augPivot[col] ?? 0) < 1e-10) {\n\t\t\treturn null;\n\t\t}\n\n\t\t// Swap rows\n\t\t[aug[col], aug[pivotRow]] = [aug[pivotRow] ?? [], aug[col] ?? []];\n\n\t\t// Scale pivot row\n\t\tconst scaledPivotRow = aug[col];\n\t\tif (scaledPivotRow === undefined) return null;\n\t\tconst pivot = scaledPivotRow[col] ?? 1;\n\n\t\tfor (let j = 0; j < 2 * n; j++) {\n\t\t\tscaledPivotRow[j] = (scaledPivotRow[j] ?? 0) / pivot;\n\t\t}\n\n\t\t// Eliminate column below pivot\n\t\tfor (let row = col + 1; row < n; row++) {\n\t\t\tconst currRow = aug[row];\n\t\t\tif (currRow === undefined) continue;\n\t\t\tconst factor = currRow[col] ?? 0;\n\n\t\t\tfor (let j = 0; j < 2 * n; j++) {\n\t\t\t\tcurrRow[j] = (currRow[j] ?? 0) - factor * (scaledPivotRow[j] ?? 0);\n\t\t\t}\n\t\t}\n\t}\n\n\t// Back substitution\n\tfor (let col = n - 1; col > 0; col--) {\n\t\tconst colRow = aug[col];\n\t\tif (colRow === undefined) return null;\n\n\t\tfor (let row = col - 1; row >= 0; row--) {\n\t\t\tconst currRow = aug[row];\n\t\t\tif (currRow === undefined) continue;\n\n\t\t\tconst factor = currRow[col] ?? 0;\n\t\t\tfor (let j = 0; j < 2 * n; j++) {\n\t\t\t\tcurrRow[j] = (currRow[j] ?? 0) - factor * (colRow[j] ?? 0);\n\t\t\t}\n\t\t}\n\t}\n\n\t// Extract inverse from augmented matrix\n\tconst inv: number[][] = [];\n\tfor (let i = 0; i < n; i++) {\n\t\tconst row: number[] = [];\n\t\tfor (let j = 0; j < n; j++) {\n\t\t\trow[j] = 0;\n\t\t}\n\t\tinv[i] = row;\n\t}\n\n\tfor (let i = 0; i < n; i++) {\n\t\tconst invRow = inv[i];\n\t\tif (invRow === undefined) continue;\n\t\tconst augRow = aug[i];\n\t\tif (augRow === undefined) continue;\n\n\t\tfor (let j = 0; j < n; j++) {\n\t\t\tinvRow[j] = augRow[n + j] ?? 0;\n\t\t}\n\t}\n\n\treturn inv;\n}\n\n/**\n * Rank paths by inverse hitting time between endpoints.\n *\n * @param graph - Source graph\n * @param paths - Paths to rank\n * @param config - Configuration options\n * @returns Ranked paths (highest inverse hitting time first)\n */\nexport function hittingTime<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tpaths: readonly ExpansionPath[],\n\tconfig?: HittingTimeConfig,\n): BaselineResult {\n\tconst {\n\t\tincludeScores = true,\n\t\tmode = \"auto\",\n\t\twalks = 1000,\n\t\tmaxSteps = 10000,\n\t\tseed = 42,\n\t} = config ?? {};\n\n\tif (paths.length === 0) {\n\t\treturn {\n\t\t\tpaths: [],\n\t\t\tmethod: \"hitting-time\",\n\t\t};\n\t}\n\n\t// Choose computation mode\n\tconst nodeCount = Array.from(graph.nodeIds()).length;\n\tconst actualMode =\n\t\tmode === \"auto\" ? (nodeCount < 100 ? \"exact\" : \"approximate\") : mode;\n\n\tconst rng = new SeededRNG(seed);\n\n\t// Score paths by inverse hitting time between endpoints\n\tconst scored: { path: ExpansionPath; score: number }[] = paths.map((path) => {\n\t\tconst source = path.nodes[0];\n\t\tconst target = path.nodes[path.nodes.length - 1];\n\n\t\tif (source === undefined || target === undefined) {\n\t\t\treturn { path, score: 0 };\n\t\t}\n\n\t\tconst ht =\n\t\t\tactualMode === \"exact\"\n\t\t\t\t? computeHittingTimeExact(graph, source, target)\n\t\t\t\t: computeHittingTimeApproximate(\n\t\t\t\t\t\tgraph,\n\t\t\t\t\t\tsource,\n\t\t\t\t\t\ttarget,\n\t\t\t\t\t\twalks,\n\t\t\t\t\t\tmaxSteps,\n\t\t\t\t\t\trng,\n\t\t\t\t\t);\n\n\t\t// Use inverse: shorter hitting time = higher score\n\t\tconst score = ht > 0 ? 1 / ht : 0;\n\t\treturn { path, score };\n\t});\n\n\t// Guard against non-finite scores before normalisation\n\tconst maxScore = Math.max(...scored.map((s) => s.score));\n\tif (!Number.isFinite(maxScore)) {\n\t\treturn {\n\t\t\tpaths: paths.map((path) => ({ ...path, score: 0 })),\n\t\t\tmethod: \"hitting-time\",\n\t\t};\n\t}\n\n\treturn normaliseAndRank(paths, scored, \"hitting-time\", includeScores);\n}\n"],"mappings":";;;;;;;;;;;;;AA4EA,SAAgB,MACf,OACA,OACA,QACc;CACd,MAAM,YAAY,YAAY,KAAK;CAEnC,MAAM,EAAE,KAAK,SAAS,UAAU,UAAU,UAAU,EAAE;CAEtD,MAAM,cAA4B,EAAE;AAEpC,MAAK,MAAM,QAAQ,OAAO;EACzB,MAAM,WAAW,oBAAoB,OAAO,MAAM,IAAI,QAAQ;AAC9D,cAAY,KAAK;GAChB,GAAG;GACH;GACA,CAAC;;AAIH,aAAY,MAAM,GAAG,MAAM,EAAE,WAAW,EAAE,SAAS;CAEnD,MAAM,UAAU,YAAY,KAAK;CAGjC,MAAM,YAAY,YAAY,KAAK,MAAM,EAAE,SAAS;CACpD,MAAM,eACL,UAAU,SAAS,IAChB,UAAU,QAAQ,GAAG,MAAM,IAAI,GAAG,EAAE,GAAG,UAAU,SACjD;CACJ,MAAM,kBAAkB,CAAC,GAAG,UAAU,CAAC,MAAM,GAAG,MAAM,IAAI,EAAE;CAC5D,MAAM,MAAM,KAAK,MAAM,gBAAgB,SAAS,EAAE;CAClD,MAAM,iBACL,gBAAgB,SAAS,IACtB,gBAAgB,SAAS,MAAM,IAC7B,gBAAgB,QAAQ,MACvB,gBAAgB,MAAM,MAAM,MAAM,gBAAgB,QAAQ,MAAM,IACnE;CACJ,MAAM,cACL,gBAAgB,SAAS,IACrB,gBAAgB,gBAAgB,SAAS,MAAM,IAChD;CACJ,MAAM,cACL,gBAAgB,SAAS,IAAK,gBAAgB,MAAM,IAAK;AAE1D,QAAO;EACN,OAAO;EACP,OAAO;GACN,aAAa,YAAY;GACzB;GACA;GACA;GACA;GACA,YAAY,UAAU;GACtB;EACD;;;;;;;;;;;;;;AA4BF,eAAsB,WACrB,OACA,OACA,QACuB;CACvB,MAAM,YAAY,YAAY,KAAK;CAEnC,MAAM,EAAE,KAAK,cAAc,UAAU,UAAU,UAAU,EAAE;CAE3D,MAAM,cAA4B,EAAE;AAEpC,MAAK,MAAM,QAAQ,OAAO;EACzB,MAAM,WAAW,MAAM,yBAAyB,OAAO,MAAM,IAAI,QAAQ;AACzE,cAAY,KAAK;GAChB,GAAG;GACH;GACA,CAAC;;AAIH,aAAY,MAAM,GAAG,MAAM,EAAE,WAAW,EAAE,SAAS;CAEnD,MAAM,UAAU,YAAY,KAAK;CAGjC,MAAM,YAAY,YAAY,KAAK,MAAM,EAAE,SAAS;CACpD,MAAM,eACL,UAAU,SAAS,IAChB,UAAU,QAAQ,GAAG,MAAM,IAAI,GAAG,EAAE,GAAG,UAAU,SACjD;CACJ,MAAM,kBAAkB,CAAC,GAAG,UAAU,CAAC,MAAM,GAAG,MAAM,IAAI,EAAE;CAC5D,MAAM,MAAM,KAAK,MAAM,gBAAgB,SAAS,EAAE;CAClD,MAAM,iBACL,gBAAgB,SAAS,IACtB,gBAAgB,SAAS,MAAM,IAC7B,gBAAgB,QAAQ,MACvB,gBAAgB,MAAM,MAAM,MAAM,gBAAgB,QAAQ,MAAM,IACnE;CACJ,MAAM,cACL,gBAAgB,SAAS,IACrB,gBAAgB,gBAAgB,SAAS,MAAM,IAChD;CACJ,MAAM,cACL,gBAAgB,SAAS,IAAK,gBAAgB,MAAM,IAAK;AAE1D,QAAO;EACN,OAAO;EACP,OAAO;GACN,aAAa,YAAY;GACzB;GACA;GACA;GACA;GACA,YAAY,UAAU;GACtB;EACD;;;;;;;;AASF,eAAe,yBACd,OACA,MACA,IACA,SACkB;CAClB,MAAM,QAAQ,KAAK;AAEnB,KAAI,MAAM,SAAS,EAClB,QAAO;CAIR,MAAM,UAAU,MAAM,QAAQ,IAC7B,MAAM,MAAM,GAAG,GAAG,CAAC,KAAK,QAAQ,MAAM;EACrC,MAAM,SAAS,MAAM,IAAI;AACzB,MAAI,WAAW,KAAA,EACd,QAAO,GAAG,OAAO,QAAQ,OAAO;AAEjC,SAAO,QAAQ,QAAQ,QAAQ;GAC9B,CACF;CAED,IAAI,YAAY;CAChB,IAAI,YAAY;AAEhB,MAAK,MAAM,UAAU,SAAS;AAC7B,eAAa,KAAK,IAAI,SAAS,OAAO;AACtC;;AAGD,KAAI,cAAc,EACjB,QAAO;CAIR,MAAM,WAAW,KAAK,IAAI,WAAW,IAAI,UAAU;AACnD,QAAO,KAAK,IAAI,SAAS,KAAK,IAAI,GAAG,SAAS,CAAC;;;;;;;AAQhD,SAAS,oBACR,OACA,MACA,IACA,SACS;CACT,MAAM,QAAQ,KAAK;AAEnB,KAAI,MAAM,SAAS,EAClB,QAAO;CAIR,IAAI,YAAY;CAChB,IAAI,YAAY;AAEhB,MAAK,IAAI,IAAI,GAAG,IAAI,MAAM,SAAS,GAAG,KAAK;EAC1C,MAAM,SAAS,MAAM;EACrB,MAAM,SAAS,MAAM,IAAI;AAEzB,MAAI,WAAW,KAAA,KAAa,WAAW,KAAA,GAAW;GACjD,MAAM,SAAS,GAAG,OAAO,QAAQ,OAAO;AACxC,gBAAa,KAAK,IAAI,SAAS,OAAO;AACtC;;;AAIF,KAAI,cAAc,EACjB,QAAO;CAIR,MAAM,WAAW,KAAK,IAAI,WAAW,IAAI,UAAU;AACnD,QAAO,KAAK,IAAI,SAAS,KAAK,IAAI,GAAG,SAAS,CAAC;;;;;;;;;;;;;;;;;ACvRhD,SAAgB,iBACf,OACA,QACA,QACA,eACiB;AACjB,KAAI,OAAO,WAAW,EACrB,QAAO;EAAE,OAAO,EAAE;EAAE;EAAQ;CAG7B,MAAM,WAAW,KAAK,IAAI,GAAG,OAAO,KAAK,MAAM,EAAE,MAAM,CAAC;AAGxD,KAAI,aAAa,EAChB,QAAO;EACN,OAAO,MAAM,KAAK,UAAU;GAAE,GAAG;GAAM,OAAO;GAAG,EAAE;EACnD;EACA;AAUF,QAAO;EAAE,OAPoB,OAC3B,KAAK,EAAE,MAAM,aAAa;GAC1B,GAAG;GACH,OAAO,gBAAgB,QAAQ,WAAW;GAC1C,EAAE,CACF,MAAM,GAAG,MAAM,EAAE,QAAQ,EAAE,MAAM;EAEX;EAAQ;;;;;;;;;;;;;;AC3BjC,SAAgB,SACf,QACA,OACA,QACiB;CACjB,MAAM,EAAE,gBAAgB,SAAS,UAAU,EAAE;AAE7C,KAAI,MAAM,WAAW,EACpB,QAAO;EACN,OAAO,EAAE;EACT,QAAQ;EACR;AAWF,QAAO,iBAAiB,OAPiC,MAAM,KAC7D,UAAU;EACV;EACA,OAAO,IAAI,KAAK,MAAM;EACtB,EACD,EAEsC,YAAY,cAAc;;;;;;;;;;;;ACvBlE,SAAgB,UACf,OACA,OACA,QACiB;CACjB,MAAM,EAAE,gBAAgB,SAAS,UAAU,EAAE;AAE7C,KAAI,MAAM,WAAW,EACpB,QAAO;EACN,OAAO,EAAE;EACT,QAAQ;EACR;AAYF,QAAO,iBAAiB,OARiC,MAAM,KAAK,SAAS;EAC5E,IAAI,YAAY;AAChB,OAAK,MAAM,UAAU,KAAK,MACzB,cAAa,MAAM,OAAO,OAAO;AAElC,SAAO;GAAE;GAAM,OAAO;GAAW;GAChC,EAEqC,cAAc,cAAc;;;;;;;;;;;;ACtBpE,SAAgB,WACf,OACA,OACA,QACiB;CACjB,MAAM,EAAE,gBAAgB,SAAS,UAAU,EAAE;AAE7C,KAAI,MAAM,WAAW,EACpB,QAAO;EACN,OAAO,EAAE;EACT,QAAQ;EACR;AA0BF,QAAO,iBAAiB,OAtBiC,MAAM,KAAK,SAAS;AAC5E,MAAI,KAAK,MAAM,SAAS,EAEvB,QAAO;GAAE;GAAM,OAAO;GAAG;EAG1B,IAAI,gBAAgB,OAAO;AAC3B,OAAK,IAAI,IAAI,GAAG,IAAI,KAAK,MAAM,SAAS,GAAG,KAAK;GAC/C,MAAM,SAAS,KAAK,MAAM;GAC1B,MAAM,SAAS,KAAK,MAAM,IAAI;AAC9B,OAAI,WAAW,KAAA,KAAa,WAAW,KAAA,EAAW;GAElD,MAAM,iBAAiB,QAAQ,OAAO,QAAQ,OAAO;AACrD,mBAAgB,KAAK,IAAI,eAAe,eAAe;;AAMxD,SAAO;GAAE;GAAM,OADd,kBAAkB,OAAO,oBAAoB,IAAI;GAC5B;GACrB,EAEqC,eAAe,cAAc;;;;;;;;;;;;ACrCrE,SAAgB,kBACf,OACA,OACA,QACiB;CACjB,MAAM,EAAE,gBAAgB,SAAS,UAAU,EAAE;AAE7C,KAAI,MAAM,WAAW,EACpB,QAAO;EACN,OAAO,EAAE;EACT,QAAQ;EACR;AA4BF,QAAO,iBAAiB,OAxBiC,MAAM,KAAK,SAAS;AAC5E,MAAI,KAAK,MAAM,SAAS,EAEvB,QAAO;GAAE;GAAM,OAAO;GAAG;EAG1B,IAAI,gBAAgB;EACpB,IAAI,YAAY;AAEhB,OAAK,IAAI,IAAI,GAAG,IAAI,KAAK,MAAM,SAAS,GAAG,KAAK;GAC/C,MAAM,SAAS,KAAK,MAAM;GAC1B,MAAM,SAAS,KAAK,MAAM,IAAI;AAC9B,OAAI,WAAW,KAAA,KAAa,WAAW,KAAA,EAAW;GAElD,MAAM,iBAAiB,QAAQ,OAAO,QAAQ,OAAO;AACrD,oBAAiB;AACjB;;AAKD,SAAO;GAAE;GAAM,OADD,YAAY,IAAI,gBAAgB,YAAY;GACpC;GACrB,EAEqC,sBAAsB,cAAc;;;;;;;;;;;;;ACtC5E,SAAS,gBACR,OACA,UAAU,KACV,YAAY,MACZ,gBAAgB,KACM;CACtB,MAAM,QAAQ,MAAM,KAAK,MAAM,SAAS,CAAC;CACzC,MAAM,IAAI,MAAM;AAEhB,KAAI,MAAM,EACT,wBAAO,IAAI,KAAK;CAIjB,MAAM,wBAAQ,IAAI,KAAqB;CACvC,MAAM,2BAAW,IAAI,KAAqB;AAC1C,MAAK,MAAM,UAAU,OAAO;AAC3B,QAAM,IAAI,QAAQ,IAAI,EAAE;AACxB,WAAS,IAAI,QAAQ,EAAE;;CAIxB,IAAI,iBAAiB;AAErB,MAAK,IAAI,YAAY,GAAG,YAAY,eAAe,aAAa;EAC/D,IAAI,YAAY;EAChB,MAAM,UAAU,iBAAiB,QAAQ;EACzC,MAAM,UAAU,iBAAiB,WAAW;AAE5C,OAAK,MAAM,UAAU,OAAO;GAE3B,IAAI,cAAc;AAElB,QAAK,MAAM,cAAc,MAAM,WAAW,QAAQ,KAAK,EAAE;IACxD,MAAM,eAAe,QAAQ,IAAI,WAAW,IAAI;IAChD,MAAM,YAAY,MAAM,OAAO,WAAW;AAC1C,QAAI,YAAY,EACf,gBAAe,eAAe;;GAKhC,MAAM,WAAW,IAAI,WAAW,IAAI,UAAU;AAC9C,WAAQ,IAAI,QAAQ,QAAQ;GAG5B,MAAM,UAAU,QAAQ,IAAI,OAAO,IAAI;AACvC,eAAY,KAAK,IAAI,WAAW,KAAK,IAAI,UAAU,QAAQ,CAAC;;AAI7D,MAAI,YAAY,UACf;AAID,mBAAiB,CAAC;AAClB,UAAQ,OAAO;;AAGhB,QAAO,iBAAiB,QAAQ;;;;;;;;;;AAWjC,SAAgB,SACf,OACA,OACA,QACiB;CACjB,MAAM,EAAE,gBAAgB,SAAS,UAAU,EAAE;AAE7C,KAAI,MAAM,WAAW,EACpB,QAAO;EACN,OAAO,EAAE;EACT,QAAQ;EACR;CAIF,MAAM,QAAQ,gBAAgB,MAAM;AAWpC,QAAO,iBAAiB,OARiC,MAAM,KAAK,SAAS;EAC5E,IAAI,QAAQ;AACZ,OAAK,MAAM,UAAU,KAAK,MACzB,UAAS,MAAM,IAAI,OAAO,IAAI;AAE/B,SAAO;GAAE;GAAM,OAAO;GAAO;GAC5B,EAEqC,YAAY,cAAc;;;;;;;;;;ACtGlE,SAAS,mBACR,OACsB;CACtB,MAAM,QAAQ,MAAM,KAAK,MAAM,SAAS,CAAC;CACzC,MAAM,8BAAc,IAAI,KAAqB;AAG7C,MAAK,MAAM,UAAU,MACpB,aAAY,IAAI,QAAQ,EAAE;AAI3B,MAAK,MAAM,UAAU,OAAO;EAE3B,MAAM,+BAAe,IAAI,KAAuB;EAChD,MAAM,2BAAW,IAAI,KAAqB;EAC1C,MAAM,wBAAQ,IAAI,KAAqB;EACvC,MAAM,QAAkB,EAAE;AAG1B,OAAK,MAAM,UAAU,OAAO;AAC3B,gBAAa,IAAI,QAAQ,EAAE,CAAC;AAC5B,YAAS,IAAI,QAAQ,GAAG;AACxB,SAAM,IAAI,QAAQ,EAAE;;AAGrB,WAAS,IAAI,QAAQ,EAAE;AACvB,QAAM,IAAI,QAAQ,EAAE;AACpB,QAAM,KAAK,OAAO;AAGlB,OAAK,MAAM,KAAK,OAAO;GACtB,MAAM,QAAQ,SAAS,IAAI,EAAE,IAAI;GACjC,MAAM,aAAa,MAAM,WAAW,EAAE;AAEtC,QAAK,MAAM,KAAK,YAAY;IAC3B,MAAM,QAAQ,SAAS,IAAI,EAAE,IAAI;AAGjC,QAAI,QAAQ,GAAG;AACd,cAAS,IAAI,GAAG,QAAQ,EAAE;AAC1B,WAAM,KAAK,EAAE;;AAId,QAAI,UAAU,QAAQ,GAAG;KACxB,MAAM,SAAS,MAAM,IAAI,EAAE,IAAI;KAC/B,MAAM,SAAS,MAAM,IAAI,EAAE,IAAI;AAC/B,WAAM,IAAI,GAAG,SAAS,OAAO;KAE7B,MAAM,QAAQ,aAAa,IAAI,EAAE,IAAI,EAAE;AACvC,WAAM,KAAK,EAAE;AACb,kBAAa,IAAI,GAAG,MAAM;;;;EAM7B,MAAM,wBAAQ,IAAI,KAAqB;AACvC,OAAK,MAAM,UAAU,MACpB,OAAM,IAAI,QAAQ,EAAE;EAIrB,MAAM,SAAS,CAAC,GAAG,MAAM,CAAC,MAAM,GAAG,MAAM;GACxC,MAAM,KAAK,SAAS,IAAI,EAAE,IAAI;AAE9B,WADW,SAAS,IAAI,EAAE,IAAI,MAClB;IACX;AAEF,OAAK,MAAM,KAAK,QAAQ;AACvB,OAAI,MAAM,OAAQ;GAElB,MAAM,SAAS,MAAM,IAAI,EAAE,IAAI;GAC/B,MAAM,SAAS,MAAM,IAAI,EAAE,IAAI;GAE/B,MAAM,QAAQ,aAAa,IAAI,EAAE,IAAI,EAAE;AACvC,QAAK,MAAM,KAAK,OAAO;IACtB,MAAM,SAAS,MAAM,IAAI,EAAE,IAAI;IAC/B,MAAM,SAAS,MAAM,IAAI,EAAE,IAAI;AAE/B,QAAI,SAAS,EACZ,OAAM,IAAI,GAAG,SAAU,SAAS,UAAW,IAAI,QAAQ;;AAIzD,OAAI,MAAM,QAAQ;IACjB,MAAM,UAAU,YAAY,IAAI,EAAE,IAAI;AACtC,gBAAY,IAAI,GAAG,UAAU,OAAO;;;;AAKvC,QAAO;;;;;;;;;;AAWR,SAAgB,YACf,OACA,OACA,QACiB;CACjB,MAAM,EAAE,gBAAgB,SAAS,UAAU,EAAE;AAE7C,KAAI,MAAM,WAAW,EACpB,QAAO;EACN,OAAO,EAAE;EACT,QAAQ;EACR;CAIF,MAAM,QAAQ,mBAAmB,MAAM;AAWvC,QAAO,iBAAiB,OARiC,MAAM,KAAK,SAAS;EAC5E,IAAI,QAAQ;AACZ,OAAK,MAAM,UAAU,KAAK,MACzB,UAAS,MAAM,IAAI,OAAO,IAAI;AAE/B,SAAO;GAAE;GAAM,OAAO;GAAO;GAC5B,EAEqC,eAAe,cAAc;;;;;;;;;;;;;;;;;AC1HrE,SAAS,YACR,OACA,QACA,QACA,IAAI,GACJ,OAAO,MACE;CACT,MAAM,QAAQ,MAAM,KAAK,MAAM,SAAS,CAAC;CACzC,MAAM,4BAAY,IAAI,KAAqB;AAC3C,OAAM,SAAS,QAAQ,QAAQ;AAC9B,YAAU,IAAI,QAAQ,IAAI;GACzB;CAEF,MAAM,IAAI,MAAM;AAChB,KAAI,MAAM,EACT,QAAO;CAGR,MAAM,YAAY,UAAU,IAAI,OAAO;CACvC,MAAM,YAAY,UAAU,IAAI,OAAO;AAEvC,KAAI,cAAc,KAAA,KAAa,cAAc,KAAA,EAC5C,QAAO;CAIR,IAAI,QAAQ,IAAI,aAAa,EAAE;AAC/B,OAAM,aAAa;CAEnB,IAAI,YAAY;AAGhB,MAAK,IAAI,QAAQ,GAAG,SAAS,GAAG,SAAS;EAExC,MAAM,YAAY,IAAI,aAAa,EAAE;AAErC,OAAK,MAAM,cAAc,OAAO;GAC/B,MAAM,SAAS,UAAU,IAAI,WAAW;AACxC,OAAI,WAAW,KAAA,EAAW;GAE1B,MAAM,aAAa,MAAM,WAAW,WAAW;AAC/C,QAAK,MAAM,eAAe,YAAY;IACrC,MAAM,OAAO,UAAU,IAAI,YAAY;AACvC,QAAI,SAAS,KAAA,EAAW;AAExB,cAAU,WAAW,UAAU,WAAW,MAAM,MAAM,SAAS;;;EAKjE,MAAM,YAAY,UAAU,cAAc;AAC1C,eAAa,KAAK,IAAI,MAAM,MAAM,GAAG;AAErC,UAAQ;;AAGT,QAAO;;;;;;;;;;AAWR,SAAgB,KACf,OACA,OACA,QACiB;CACjB,MAAM,EAAE,gBAAgB,SAAS,UAAU,EAAE;AAE7C,KAAI,MAAM,WAAW,EACpB,QAAO;EACN,OAAO,EAAE;EACT,QAAQ;EACR;AAgBF,QAAO,iBAAiB,OAZiC,MAAM,KAAK,SAAS;EAC5E,MAAM,SAAS,KAAK,MAAM;EAC1B,MAAM,SAAS,KAAK,MAAM,KAAK,MAAM,SAAS;AAE9C,MAAI,WAAW,KAAA,KAAa,WAAW,KAAA,EACtC,QAAO;GAAE;GAAM,OAAO;GAAG;AAI1B,SAAO;GAAE;GAAM,OADG,YAAY,OAAO,QAAQ,OAAO;GACnB;GAChC,EAEqC,QAAQ,cAAc;;;;;;;;;;;;;;;AChG9D,SAAS,uBACR,OACA,QACA,QACA,IAAI,IACK;CACT,MAAM,QAAQ,MAAM,KAAK,MAAM,SAAS,CAAC;CACzC,MAAM,4BAAY,IAAI,KAAqB;AAC3C,OAAM,SAAS,QAAQ,QAAQ;AAC9B,YAAU,IAAI,QAAQ,IAAI;GACzB;CAEF,MAAM,IAAI,MAAM;AAChB,KAAI,MAAM,EACT,QAAO;CAGR,MAAM,YAAY,UAAU,IAAI,OAAO;CACvC,MAAM,YAAY,UAAU,IAAI,OAAO;AAEvC,KAAI,cAAc,KAAA,KAAa,cAAc,KAAA,EAC5C,QAAO;CAIR,IAAI,QAAQ,IAAI,aAAa,EAAE;AAC/B,OAAM,aAAa;CAGnB,IAAI,YAAY,MAAM,cAAc;CAEpC,IAAI,YAAY;AAGhB,MAAK,IAAI,QAAQ,GAAG,SAAS,GAAG,SAAS;EAExC,MAAM,YAAY,IAAI,aAAa,EAAE;AAErC,OAAK,MAAM,YAAY,OAAO;GAC7B,MAAM,UAAU,UAAU,IAAI,SAAS;AACvC,OAAI,YAAY,KAAA,EAAW;GAE3B,MAAM,aAAa,MAAM,WAAW,SAAS;AAC7C,QAAK,MAAM,YAAY,YAAY;IAClC,MAAM,QAAQ,UAAU,IAAI,SAAS;AACrC,QAAI,UAAU,KAAA,EAAW;AAEzB,cAAU,YAAY,UAAU,YAAY,MAAM,MAAM,UAAU;;;AAIpE,eAAa;AAGb,gBAAc,UAAU,cAAc,KAAK;AAE3C,UAAQ;;AAGT,QAAO;;;;;;;;;;AAWR,SAAgB,gBACf,OACA,OACA,QACiB;CACjB,MAAM,EAAE,gBAAgB,SAAS,UAAU,EAAE;AAE7C,KAAI,MAAM,WAAW,EACpB,QAAO;EACN,OAAO,EAAE;EACT,QAAQ;EACR;AAgBF,QAAO,iBAAiB,OAZiC,MAAM,KAAK,SAAS;EAC5E,MAAM,SAAS,KAAK,MAAM;EAC1B,MAAM,SAAS,KAAK,MAAM,KAAK,MAAM,SAAS;AAE9C,MAAI,WAAW,KAAA,KAAa,WAAW,KAAA,EACtC,QAAO;GAAE;GAAM,OAAO;GAAG;AAI1B,SAAO;GAAE;GAAM,OADG,uBAAuB,OAAO,QAAQ,OAAO;GAC9B;GAChC,EAEqC,mBAAmB,cAAc;;;;;;;;;;;;;;;ACjGzE,SAAS,kBACR,OACA,QACA,QACS;CACT,MAAM,QAAQ,MAAM,KAAK,MAAM,SAAS,CAAC;CACzC,MAAM,4BAAY,IAAI,KAAqB;AAC3C,OAAM,SAAS,QAAQ,QAAQ;AAC9B,YAAU,IAAI,QAAQ,IAAI;GACzB;CAEF,MAAM,IAAI,MAAM;AAEhB,KAAI,MAAM,KAAK,IAAI,IAClB,OAAM,IAAI,MACT,wDAAwD,OAAO,EAAE,CAAC,wBAClE;CAGF,MAAM,YAAY,UAAU,IAAI,OAAO;CACvC,MAAM,YAAY,UAAU,IAAI,OAAO;AAEvC,KAAI,cAAc,KAAA,KAAa,cAAc,KAAA,EAC5C,QAAO;CAIR,MAAM,IAAgB,MAAM,KAAK,EAAE,QAAQ,GAAG,QAC7C,MAAM,KAAK,EAAE,QAAQ,GAAG,QAAQ,EAAE,CAClC;AAED,MAAK,IAAI,IAAI,GAAG,IAAI,GAAG,KAAK;EAC3B,MAAM,SAAS,MAAM;AACrB,MAAI,WAAW,KAAA,EAAW;EAE1B,MAAM,SAAS,MAAM,OAAO,OAAO;EACnC,MAAM,MAAM,EAAE;AACd,MAAI,QAAQ,KAAA,EACX,KAAI,KAAK;EAGV,MAAM,aAAa,MAAM,WAAW,OAAO;AAC3C,OAAK,MAAM,eAAe,YAAY;GACrC,MAAM,IAAI,UAAU,IAAI,YAAY;AACpC,OAAI,MAAM,KAAA,KAAa,QAAQ,KAAA,EAC9B,KAAI,KAAK;;;CAOZ,MAAM,QAAQ,KAAK,EAAE;CAGrB,MAAM,cACJ,MAAM,aAAa,cAAc,MACjC,MAAM,aAAa,cAAc,KAClC,KAAK,MAAM,aAAa,cAAc;AAGvC,QAAO,KAAK,IAAI,YAAY,MAAM;;;;;;;;;AAUnC,SAAS,KAAK,GAA2B;CACxC,MAAM,IAAI,EAAE;AACZ,KAAI,MAAM,EAAG,QAAO,EAAE;CAGtB,MAAM,IAAI,EAAE,KAAK,QAAQ,CAAC,GAAG,IAAI,CAAC;CAGlC,MAAM,UAAU;AAChB,MAAK,IAAI,IAAI,GAAG,IAAI,GAAG,KAAK;EAC3B,MAAM,MAAM,EAAE;AACd,MAAI,QAAQ,KAAA,EACX,KAAI,MAAM,IAAI,MAAM,KAAK;;AAO3B,QAFa,gBAAgB,EAAE;;;;;;;;AAWhC,SAAS,gBAAgB,GAA2B;CACnD,MAAM,IAAI,EAAE;CAGZ,MAAM,MAAkB,EAAE,KAAK,KAAK,MAAM;EACzC,MAAM,WAAqB,MAAM,KAAK,EAAE,QAAQ,GAAG,GAAG,GAAG,MACxD,MAAM,IAAI,IAAI,EACd;AAED,SAD2B,CAAC,GAAG,KAAK,GAAG,SAAS;GAE/C;AAGF,MAAK,IAAI,MAAM,GAAG,MAAM,GAAG,OAAO;EAEjC,IAAI,SAAS;AACb,OAAK,IAAI,MAAM,MAAM,GAAG,MAAM,GAAG,OAAO;GACvC,MAAM,aAAa,IAAI;GACvB,MAAM,YAAY,IAAI;AACtB,OACC,eAAe,KAAA,KACf,cAAc,KAAA,KACd,KAAK,IAAI,WAAW,QAAQ,EAAE,GAAG,KAAK,IAAI,UAAU,QAAQ,EAAE,CAE9D,UAAS;;EAKX,MAAM,aAAa,IAAI;EACvB,MAAM,YAAY,IAAI;AACtB,MAAI,eAAe,KAAA,KAAa,cAAc,KAAA,GAAW;AACxD,OAAI,OAAO;AACX,OAAI,UAAU;;EAIf,MAAM,WAAW,IAAI;EACrB,MAAM,QAAQ,WAAW;AACzB,MAAI,UAAU,KAAA,KAAa,KAAK,IAAI,MAAM,GAAG,MAC5C;AAGD,MAAI,aAAa,KAAA,EAChB,MAAK,IAAI,IAAI,KAAK,IAAI,IAAI,GAAG,IAC5B,UAAS,MAAM,SAAS,MAAM,KAAK;AAKrC,OAAK,IAAI,MAAM,GAAG,MAAM,GAAG,OAAO;AACjC,OAAI,QAAQ,IAAK;GAEjB,MAAM,iBAAiB,IAAI;GAC3B,MAAM,SAAS,iBAAiB,QAAQ;AACxC,OAAI,mBAAmB,KAAA,KAAa,aAAa,KAAA,EAChD,MAAK,IAAI,IAAI,KAAK,IAAI,IAAI,GAAG,IAC5B,gBAAe,MACb,eAAe,MAAM,KAAK,UAAU,SAAS,MAAM;;;CAOzD,MAAM,OAAmB,EAAE;AAC3B,MAAK,IAAI,IAAI,GAAG,IAAI,GAAG,IAEtB,MAAK,MADO,IAAI,IACA,MAAM,EAAE,IAAI,EAAE,EAAE,KAAK,MAAM,EAAE;AAG9C,QAAO;;;;;;;;;;AAWR,SAAgB,mBACf,OACA,OACA,QACiB;CACjB,MAAM,EAAE,gBAAgB,SAAS,UAAU,EAAE;AAE7C,KAAI,MAAM,WAAW,EACpB,QAAO;EACN,OAAO,EAAE;EACT,QAAQ;EACR;CAIF,MAAM,YAAY,MAAM,KAAK,MAAM,SAAS,CAAC,CAAC;AAC9C,KAAI,YAAY,IACf,OAAM,IAAI,MACT,uCAAuC,OAAO,UAAU,CAAC,+EACzD;AAkBF,QAAO,iBAAiB,OAdiC,MAAM,KAAK,SAAS;EAC5E,MAAM,SAAS,KAAK,MAAM;EAC1B,MAAM,SAAS,KAAK,MAAM,KAAK,MAAM,SAAS;AAE9C,MAAI,WAAW,KAAA,KAAa,WAAW,KAAA,EACtC,QAAO;GAAE;GAAM,OAAO;GAAG;AAM1B,SAAO;GAAE;GAAM,OADD,IAFK,kBAAkB,OAAO,QAAQ,OAAO;GAGrC;GACrB,EAEqC,uBAAuB,cAAc;;;;;;;;;;;;ACrN7E,SAAS,aAAa,OAAe,OAAO,GAAW;CACtD,IAAI,IAAI;AACR,MAAK,IAAI,IAAI,GAAG,IAAI,MAAM,QAAQ,KAAK;AACtC,MAAI,KAAK,KAAK,IAAI,MAAM,WAAW,EAAE,EAAE,WAAW;AAElD,OAAK,MAAM;;AAGZ,SAAQ,MAAM,KAAK;;;;;;;;;;AAWpB,SAAgB,cACf,QACA,OACA,QACiB;CACjB,MAAM,EAAE,gBAAgB,MAAM,OAAO,MAAM,UAAU,EAAE;AAEvD,KAAI,MAAM,WAAW,EACpB,QAAO;EACN,OAAO,EAAE;EACT,QAAQ;EACR;AAUF,QAAO,iBAAiB,OANiC,MAAM,KAAK,SAAS;AAG5E,SAAO;GAAE;GAAM,OADD,aADG,KAAK,MAAM,KAAK,IAAI,EACA,KAAK;GACpB;GACrB,EAEqC,UAAU,cAAc;;;;;;;;AChChE,IAAM,YAAN,MAAgB;CACf;CAEA,YAAY,MAAc;AACzB,OAAK,QAAQ;;;;;CAMd,OAAe;AAEd,OAAK,QAAS,KAAK,QAAQ,aAAa,QAAS;AACjD,SAAO,KAAK,QAAQ;;;;;;;;;;;;;;AAetB,SAAS,8BACR,OACA,QACA,QACA,OACA,UACA,KACS;AACT,KAAI,WAAW,OACd,QAAO;CAGR,IAAI,aAAa;CACjB,IAAI,kBAAkB;AAEtB,MAAK,IAAI,IAAI,GAAG,IAAI,OAAO,KAAK;EAC/B,IAAI,UAAkB;EACtB,IAAI,QAAQ;AAEZ,SAAO,YAAY,UAAU,QAAQ,UAAU;GAC9C,MAAM,aAAa,MAAM,KAAK,MAAM,WAAW,QAAQ,CAAC;AACxD,OAAI,WAAW,WAAW,EAEzB;GAKD,MAAM,WAAW,WADD,KAAK,MAAM,IAAI,MAAM,GAAG,WAAW,OAAO;AAE1D,OAAI,aAAa,KAAA,EAChB;AAED,aAAU;AACV;;AAGD,MAAI,YAAY,QAAQ;AACvB,iBAAc;AACd;;;AAKF,KAAI,kBAAkB,EACrB,QAAO,aAAa;AAIrB,QAAO;;;;;;;;;;;;;AAcR,SAAS,wBACR,OACA,QACA,QACS;AACT,KAAI,WAAW,OACd,QAAO;CAGR,MAAM,QAAQ,MAAM,KAAK,MAAM,SAAS,CAAC;CACzC,MAAM,4BAAY,IAAI,KAAqB;AAC3C,OAAM,SAAS,QAAQ,QAAQ;AAC9B,YAAU,IAAI,QAAQ,IAAI;GACzB;CAEF,MAAM,IAAI,MAAM;CAChB,MAAM,YAAY,UAAU,IAAI,OAAO;CACvC,MAAM,YAAY,UAAU,IAAI,OAAO;AAEvC,KAAI,cAAc,KAAA,KAAa,cAAc,KAAA,EAC5C,QAAO;CAIR,MAAM,IAAgB,EAAE;AACxB,MAAK,IAAI,IAAI,GAAG,IAAI,GAAG,KAAK;EAC3B,MAAM,MAAgB,EAAE;AACxB,OAAK,IAAI,IAAI,GAAG,IAAI,GAAG,IACtB,KAAI,KAAK;AAEV,IAAE,KAAK;;AAGR,MAAK,MAAM,UAAU,OAAO;EAC3B,MAAM,MAAM,UAAU,IAAI,OAAO;AACjC,MAAI,QAAQ,KAAA,EAAW;EAEvB,MAAM,OAAO,EAAE;AACf,MAAI,SAAS,KAAA,EAAW;AAExB,MAAI,QAAQ,UAEX,MAAK,OAAO;OACN;GACN,MAAM,aAAa,MAAM,KAAK,MAAM,WAAW,OAAO,CAAC;GACvD,MAAM,SAAS,WAAW;AAE1B,OAAI,SAAS,EACZ,MAAK,MAAM,eAAe,YAAY;IACrC,MAAM,OAAO,UAAU,IAAI,YAAY;AACvC,QAAI,SAAS,KAAA,EACZ,MAAK,QAAQ,IAAI;;;;CAStB,MAAM,mBAA6B,EAAE;AAErC,MAAK,IAAI,IAAI,GAAG,IAAI,GAAG,IACtB,KAAI,MAAM,UACT,kBAAiB,KAAK,EAAE;CAI1B,MAAM,IAAI,iBAAiB;CAC3B,MAAM,IAAgB,EAAE;AACxB,MAAK,IAAI,IAAI,GAAG,IAAI,GAAG,KAAK;EAC3B,MAAM,MAAgB,EAAE;AACxB,OAAK,IAAI,IAAI,GAAG,IAAI,GAAG,IACtB,KAAI,KAAK;AAEV,IAAE,KAAK;;AAGR,MAAK,IAAI,IAAI,GAAG,IAAI,GAAG,KAAK;EAC3B,MAAM,OAAO,EAAE;AACf,MAAI,SAAS,KAAA,EAAW;EACxB,MAAM,QAAQ,iBAAiB;AAC/B,MAAI,UAAU,KAAA,EAAW;EACzB,MAAM,OAAO,EAAE;AACf,MAAI,SAAS,KAAA,EAAW;AAExB,OAAK,IAAI,IAAI,GAAG,IAAI,GAAG,KAAK;GAC3B,MAAM,QAAQ,iBAAiB;AAC/B,OAAI,UAAU,KAAA,EAAW;AACzB,QAAK,KAAK,KAAK,UAAU;;;CAK3B,MAAM,MAAkB,EAAE;AAC1B,MAAK,IAAI,IAAI,GAAG,IAAI,GAAG,KAAK;EAC3B,MAAM,MAAgB,EAAE;AACxB,OAAK,IAAI,IAAI,GAAG,IAAI,GAAG,IACtB,KAAI,KAAK,MAAM,IAAI,IAAI;AAExB,MAAI,KAAK;;AAGV,MAAK,IAAI,IAAI,GAAG,IAAI,GAAG,KAAK;EAC3B,MAAM,SAAS,IAAI;AACnB,MAAI,WAAW,KAAA,EAAW;EAC1B,MAAM,OAAO,EAAE;AAEf,OAAK,IAAI,IAAI,GAAG,IAAI,GAAG,KAAK;GAC3B,MAAM,OAAO,OAAO,MAAM;AAC1B,UAAO,MAAM,MAAM,IAAI,IAAI,KAAK;;;CAKlC,MAAM,IAAI,aAAa,IAAI;AAE3B,KAAI,MAAM,KAET,QAAO;CAIR,MAAM,qBAAqB,iBAAiB,QAAQ,UAAU;AAC9D,KAAI,qBAAqB,EACxB,QAAO;CAGR,IAAI,cAAc;CAClB,MAAM,MAAM,EAAE;AACd,KAAI,QAAQ,KAAA,EACX,MAAK,MAAM,OAAO,IACjB,gBAAe;AAIjB,QAAO;;;;;;;;AASR,SAAS,aAAa,QAAuC;CAC5D,MAAM,IAAI,OAAO;CACjB,MAAM,MAAkB,EAAE;AAG1B,MAAK,IAAI,IAAI,GAAG,IAAI,GAAG,KAAK;EAC3B,MAAM,MAAgB,EAAE;EACxB,MAAM,SAAS,OAAO;AAEtB,OAAK,IAAI,IAAI,GAAG,IAAI,GAAG,IACtB,KAAI,KAAK,SAAS,MAAM;AAEzB,OAAK,IAAI,IAAI,GAAG,IAAI,GAAG,IACtB,KAAI,IAAI,KAAK,MAAM,IAAI,IAAI;AAE5B,MAAI,KAAK;;AAIV,MAAK,IAAI,MAAM,GAAG,MAAM,GAAG,OAAO;EAEjC,IAAI,WAAW;EACf,MAAM,WAAW,IAAI;AACrB,MAAI,aAAa,KAAA,EAAW,QAAO;AAEnC,OAAK,IAAI,MAAM,MAAM,GAAG,MAAM,GAAG,OAAO;GACvC,MAAM,aAAa,IAAI,OAAO,QAAQ;GACtC,MAAM,cAAc,SAAS,QAAQ;AACrC,OAAI,KAAK,IAAI,WAAW,GAAG,KAAK,IAAI,YAAY,CAC/C,YAAW;;EAKb,MAAM,WAAW,IAAI;AACrB,MAAI,aAAa,KAAA,KAAa,KAAK,IAAI,SAAS,QAAQ,EAAE,GAAG,MAC5D,QAAO;AAIR,GAAC,IAAI,MAAM,IAAI,aAAa,CAAC,IAAI,aAAa,EAAE,EAAE,IAAI,QAAQ,EAAE,CAAC;EAGjE,MAAM,iBAAiB,IAAI;AAC3B,MAAI,mBAAmB,KAAA,EAAW,QAAO;EACzC,MAAM,QAAQ,eAAe,QAAQ;AAErC,OAAK,IAAI,IAAI,GAAG,IAAI,IAAI,GAAG,IAC1B,gBAAe,MAAM,eAAe,MAAM,KAAK;AAIhD,OAAK,IAAI,MAAM,MAAM,GAAG,MAAM,GAAG,OAAO;GACvC,MAAM,UAAU,IAAI;AACpB,OAAI,YAAY,KAAA,EAAW;GAC3B,MAAM,SAAS,QAAQ,QAAQ;AAE/B,QAAK,IAAI,IAAI,GAAG,IAAI,IAAI,GAAG,IAC1B,SAAQ,MAAM,QAAQ,MAAM,KAAK,UAAU,eAAe,MAAM;;;AAMnE,MAAK,IAAI,MAAM,IAAI,GAAG,MAAM,GAAG,OAAO;EACrC,MAAM,SAAS,IAAI;AACnB,MAAI,WAAW,KAAA,EAAW,QAAO;AAEjC,OAAK,IAAI,MAAM,MAAM,GAAG,OAAO,GAAG,OAAO;GACxC,MAAM,UAAU,IAAI;AACpB,OAAI,YAAY,KAAA,EAAW;GAE3B,MAAM,SAAS,QAAQ,QAAQ;AAC/B,QAAK,IAAI,IAAI,GAAG,IAAI,IAAI,GAAG,IAC1B,SAAQ,MAAM,QAAQ,MAAM,KAAK,UAAU,OAAO,MAAM;;;CAM3D,MAAM,MAAkB,EAAE;AAC1B,MAAK,IAAI,IAAI,GAAG,IAAI,GAAG,KAAK;EAC3B,MAAM,MAAgB,EAAE;AACxB,OAAK,IAAI,IAAI,GAAG,IAAI,GAAG,IACtB,KAAI,KAAK;AAEV,MAAI,KAAK;;AAGV,MAAK,IAAI,IAAI,GAAG,IAAI,GAAG,KAAK;EAC3B,MAAM,SAAS,IAAI;AACnB,MAAI,WAAW,KAAA,EAAW;EAC1B,MAAM,SAAS,IAAI;AACnB,MAAI,WAAW,KAAA,EAAW;AAE1B,OAAK,IAAI,IAAI,GAAG,IAAI,GAAG,IACtB,QAAO,KAAK,OAAO,IAAI,MAAM;;AAI/B,QAAO;;;;;;;;;;AAWR,SAAgB,YACf,OACA,OACA,QACiB;CACjB,MAAM,EACL,gBAAgB,MAChB,OAAO,QACP,QAAQ,KACR,WAAW,KACX,OAAO,OACJ,UAAU,EAAE;AAEhB,KAAI,MAAM,WAAW,EACpB,QAAO;EACN,OAAO,EAAE;EACT,QAAQ;EACR;CAIF,MAAM,YAAY,MAAM,KAAK,MAAM,SAAS,CAAC,CAAC;CAC9C,MAAM,aACL,SAAS,SAAU,YAAY,MAAM,UAAU,gBAAiB;CAEjE,MAAM,MAAM,IAAI,UAAU,KAAK;CAG/B,MAAM,SAAmD,MAAM,KAAK,SAAS;EAC5E,MAAM,SAAS,KAAK,MAAM;EAC1B,MAAM,SAAS,KAAK,MAAM,KAAK,MAAM,SAAS;AAE9C,MAAI,WAAW,KAAA,KAAa,WAAW,KAAA,EACtC,QAAO;GAAE;GAAM,OAAO;GAAG;EAG1B,MAAM,KACL,eAAe,UACZ,wBAAwB,OAAO,QAAQ,OAAO,GAC9C,8BACA,OACA,QACA,QACA,OACA,UACA,IACA;AAIJ,SAAO;GAAE;GAAM,OADD,KAAK,IAAI,IAAI,KAAK;GACV;GACrB;CAGF,MAAM,WAAW,KAAK,IAAI,GAAG,OAAO,KAAK,MAAM,EAAE,MAAM,CAAC;AACxD,KAAI,CAAC,OAAO,SAAS,SAAS,CAC7B,QAAO;EACN,OAAO,MAAM,KAAK,UAAU;GAAE,GAAG;GAAM,OAAO;GAAG,EAAE;EACnD,QAAQ;EACR;AAGF,QAAO,iBAAiB,OAAO,QAAQ,gBAAgB,cAAc"}