graphwise 1.8.0 → 1.9.0

package/dist/ranking/mi/index.cjs.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.cjs","names":[],"sources":["../../../src/ranking/mi/adamic-adar.ts","../../../src/ranking/mi/cosine.ts","../../../src/ranking/mi/sorensen.ts","../../../src/ranking/mi/resource-allocation.ts","../../../src/ranking/mi/overlap-coefficient.ts","../../../src/ranking/mi/hub-promoted.ts","../../../src/ranking/mi/scale.ts","../../../src/ranking/mi/skew.ts","../../../src/ranking/mi/span.ts","../../../src/ranking/mi/etch.ts","../../../src/ranking/mi/notch.ts","../../../src/ranking/mi/adaptive.ts"],"sourcesContent":["/**\n * Adamic-Adar index for edge salience.\n *\n * Sum of inverse log degrees of common neighbours:\n * MI(u,v) = Σ_{z ∈ N(u) ∩ N(v)} 1 / log(deg(z) + 1)\n *\n * Range: [0, ∞) - higher values indicate stronger association\n * Normalised to [0, 1] by dividing by max possible value.\n */\n\nimport type { NodeId, NodeData, EdgeData, ReadableGraph } from \"../../graph\";\nimport type { AsyncReadableGraph } from \"../../graph/async-interfaces\";\nimport { neighbourSet, neighbourIntersection } from \"../../utils\";\nimport { collectAsyncIterable } from \"../../async/utils\";\nimport type { MIConfig } from \"./types\";\n\n/**\n * Compute Adamic-Adar index between neighbourhoods of two nodes.\n *\n * @param graph - Source graph\n * @param source - Source node ID\n * @param target - Target node ID\n * @param config - Optional configuration\n * @returns Adamic-Adar index (normalised to [0, 1] if configured)\n */\nexport function adamicAdar<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tsource: NodeId,\n\ttarget: NodeId,\n\tconfig?: MIConfig,\n): number {\n\tconst { epsilon = 1e-10, normalise = true } = config ?? {};\n\n\t// Get neighbourhoods, excluding opposite endpoint\n\tconst sourceNeighbours = neighbourSet(graph, source, target);\n\tconst targetNeighbours = neighbourSet(graph, target, source);\n\n\t// Compute common neighbours\n\tconst commonNeighbours = neighbourIntersection(\n\t\tsourceNeighbours,\n\t\ttargetNeighbours,\n\t);\n\n\t// Sum inverse log degrees of common neighbours\n\tlet score = 0;\n\tfor (const neighbour of commonNeighbours) {\n\t\tconst degree = graph.degree(neighbour);\n\t\tscore += 1 / Math.log(degree + 1);\n\t}\n\n\t// Normalise to [0, 1] if requested\n\tif (normalise && commonNeighbours.size > 0) {\n\t\t// Max possible is when all common neighbours have minimum degree (1)\n\t\t// 1 / log(1 + 1) = 1 / log(2)\n\t\tconst maxScore = commonNeighbours.size / Math.log(2);\n\t\tscore = score / maxScore;\n\t}\n\n\t// Apply epsilon floor for numerical stability\n\treturn Math.max(epsilon, score);\n}\n\n/**\n * Async variant of Adamic-Adar index for use with async graph data sources.\n *\n * Fetches both neighbourhoods concurrently, then fetches degree for each common\n * neighbour to compute the inverse-log-degree weighted sum.\n */\nexport async function adamicAdarAsync<N extends NodeData, E extends EdgeData>(\n\tgraph: AsyncReadableGraph<N, E>,\n\tsource: NodeId,\n\ttarget: NodeId,\n\tconfig?: MIConfig,\n): Promise<number> {\n\tconst { epsilon = 1e-10, normalise = true } = config ?? {};\n\n\t// Fetch both neighbourhoods in parallel\n\tconst [sourceArr, targetArr] = await Promise.all([\n\t\tcollectAsyncIterable(graph.neighbours(source)),\n\t\tcollectAsyncIterable(graph.neighbours(target)),\n\t]);\n\n\tconst srcSet = new Set(sourceArr.filter((n) => n !== target));\n\tconst tgtSet = new Set(targetArr.filter((n) => n !== source));\n\n\t// Find common neighbours\n\tconst commonNeighbours: NodeId[] = [];\n\tfor (const n of srcSet) {\n\t\tif (tgtSet.has(n)) commonNeighbours.push(n);\n\t}\n\n\tif (commonNeighbours.length === 0) {\n\t\treturn epsilon;\n\t}\n\n\t// Fetch degrees of all common neighbours in parallel\n\tconst degrees = await Promise.all(\n\t\tcommonNeighbours.map((n) => graph.degree(n)),\n\t);\n\n\t// Sum inverse log degrees\n\tlet score = 0;\n\tfor (const degree of degrees) {\n\t\tscore += 1 / Math.log(degree + 1);\n\t}\n\n\t// Normalise to [0, 1] if requested\n\tif (normalise) {\n\t\t// Max possible is when all common neighbours have minimum degree (1)\n\t\t// 1 / log(1 + 1) = 1 / log(2)\n\t\tconst maxScore = commonNeighbours.length / Math.log(2);\n\t\tscore = score / maxScore;\n\t}\n\n\treturn Math.max(epsilon, score);\n}\n","/**\n * Cosine similarity for edge salience.\n *\n * Measures similarity between neighbourhoods using vector cosine:\n * MI(u,v) = |N(u) ∩ N(v)| / (√|N(u)| × √|N(v)|)\n *\n * Range: [0, 1]\n * - 0: No overlap or one node has no neighbours\n * - 1: Identical neighbourhoods of equal size\n */\n\nimport type { NodeId, NodeData, EdgeData, ReadableGraph } from \"../../graph\";\nimport type { AsyncReadableGraph } from \"../../graph/async-interfaces\";\nimport { neighbourSet, neighbourOverlap } from \"../../utils\";\nimport { collectAsyncIterable } from \"../../async/utils\";\nimport type { MIConfig } from \"./types\";\n\n/**\n * Compute cosine similarity between neighbourhoods of two nodes.\n *\n * @param graph - Source graph\n * @param source - Source node ID\n * @param target - Target node ID\n * @param config - Optional configuration\n * @returns Cosine similarity in [0, 1]\n */\nexport function cosine<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tsource: NodeId,\n\ttarget: NodeId,\n\tconfig?: MIConfig,\n): number {\n\tconst { epsilon = 1e-10 } = config ?? {};\n\n\t// Get neighbourhoods, excluding opposite endpoint\n\tconst sourceNeighbours = neighbourSet(graph, source, target);\n\tconst targetNeighbours = neighbourSet(graph, target, source);\n\n\t// Compute intersection size\n\tconst { intersection } = neighbourOverlap(sourceNeighbours, targetNeighbours);\n\n\t// Compute denominator: √|N(u)| × √|N(v)|\n\tconst denominator =\n\t\tMath.sqrt(sourceNeighbours.size) * Math.sqrt(targetNeighbours.size);\n\n\t// Avoid division by zero\n\tif (denominator === 0) {\n\t\treturn 0;\n\t}\n\n\tconst score = intersection / denominator;\n\n\t// Apply epsilon floor for numerical stability\n\treturn Math.max(epsilon, score);\n}\n\n/**\n * Async variant of cosine similarity for use with async graph data sources.\n *\n * Fetches both neighbourhoods concurrently, then applies the same formula.\n */\nexport async function cosineAsync<N extends NodeData, E extends EdgeData>(\n\tgraph: AsyncReadableGraph<N, E>,\n\tsource: NodeId,\n\ttarget: NodeId,\n\tconfig?: MIConfig,\n): Promise<number> {\n\tconst { epsilon = 1e-10 } = config ?? {};\n\n\t// Fetch both neighbourhoods in parallel\n\tconst [sourceArr, targetArr] = await Promise.all([\n\t\tcollectAsyncIterable(graph.neighbours(source)),\n\t\tcollectAsyncIterable(graph.neighbours(target)),\n\t]);\n\n\tconst srcSet = new Set(sourceArr.filter((n) => n !== target));\n\tconst tgtSet = new Set(targetArr.filter((n) => n !== source));\n\n\t// Compute intersection size\n\tlet intersection = 0;\n\tfor (const n of srcSet) {\n\t\tif (tgtSet.has(n)) intersection++;\n\t}\n\n\t// Compute denominator: √|N(u)| × √|N(v)|\n\tconst denominator = Math.sqrt(srcSet.size) * Math.sqrt(tgtSet.size);\n\n\t// Avoid division by zero\n\tif (denominator === 0) {\n\t\treturn 0;\n\t}\n\n\tconst score = intersection / denominator;\n\n\treturn Math.max(epsilon, score);\n}\n","/**\n * Sorensen-Dice coefficient for edge salience.\n *\n * Harmonic mean of overlap:\n * MI(u,v) = 2|N(u) ∩ N(v)| / (|N(u)| + |N(v)|)\n *\n * Range: [0, 1]\n * - 0: No shared neighbours\n * - 1: Identical neighbourhoods\n */\n\nimport type { NodeId, NodeData, EdgeData, ReadableGraph } from \"../../graph\";\nimport type { AsyncReadableGraph } from \"../../graph/async-interfaces\";\nimport { neighbourSet, neighbourOverlap } from \"../../utils\";\nimport { collectAsyncIterable } from \"../../async/utils\";\nimport type { MIConfig } from \"./types\";\n\n/**\n * Compute Sorensen-Dice similarity between neighbourhoods of two nodes.\n *\n * @param graph - Source graph\n * @param source - Source node ID\n * @param target - Target node ID\n * @param config - Optional configuration\n * @returns Sorensen-Dice coefficient in [0, 1]\n */\nexport function sorensen<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tsource: NodeId,\n\ttarget: NodeId,\n\tconfig?: MIConfig,\n): number {\n\tconst { epsilon = 1e-10 } = config ?? {};\n\n\t// Get neighbourhoods, excluding opposite endpoint\n\tconst sourceNeighbours = neighbourSet(graph, source, target);\n\tconst targetNeighbours = neighbourSet(graph, target, source);\n\n\t// Compute intersection size\n\tconst { intersection } = neighbourOverlap(sourceNeighbours, targetNeighbours);\n\n\t// Compute denominator: |N(u)| + |N(v)|\n\tconst denominator = sourceNeighbours.size + targetNeighbours.size;\n\n\t// Avoid division by zero\n\tif (denominator === 0) {\n\t\treturn 0;\n\t}\n\n\tconst score = (2 * intersection) / denominator;\n\n\t// Apply epsilon floor for numerical stability\n\treturn Math.max(epsilon, score);\n}\n\n/**\n * Async variant of Sorensen-Dice coefficient for use with async graph data sources.\n *\n * Fetches both neighbourhoods concurrently, then applies the same formula.\n */\nexport async function sorensenAsync<N extends NodeData, E extends EdgeData>(\n\tgraph: AsyncReadableGraph<N, E>,\n\tsource: NodeId,\n\ttarget: NodeId,\n\tconfig?: MIConfig,\n): Promise<number> {\n\tconst { epsilon = 1e-10 } = config ?? {};\n\n\t// Fetch both neighbourhoods in parallel\n\tconst [sourceArr, targetArr] = await Promise.all([\n\t\tcollectAsyncIterable(graph.neighbours(source)),\n\t\tcollectAsyncIterable(graph.neighbours(target)),\n\t]);\n\n\tconst srcSet = new Set(sourceArr.filter((n) => n !== target));\n\tconst tgtSet = new Set(targetArr.filter((n) => n !== source));\n\n\t// Compute intersection size\n\tlet intersection = 0;\n\tfor (const n of srcSet) {\n\t\tif (tgtSet.has(n)) intersection++;\n\t}\n\n\t// Compute denominator: |N(u)| + |N(v)|\n\tconst denominator = srcSet.size + tgtSet.size;\n\n\t// Avoid division by zero\n\tif (denominator === 0) {\n\t\treturn 0;\n\t}\n\n\tconst score = (2 * intersection) / denominator;\n\n\treturn Math.max(epsilon, score);\n}\n","/**\n * Resource Allocation index for edge salience.\n *\n * Sum of inverse degrees of common neighbours:\n * MI(u,v) = Σ_{z ∈ N(u) ∩ N(v)} 1 / deg(z)\n *\n * Range: [0, ∞)\n * Assumption: shared neighbours with low degree are more informative than hubs.\n */\n\nimport type { NodeId, NodeData, EdgeData, ReadableGraph } from \"../../graph\";\nimport type { AsyncReadableGraph } from \"../../graph/async-interfaces\";\nimport { neighbourSet, neighbourIntersection } from \"../../utils\";\nimport { collectAsyncIterable } from \"../../async/utils\";\nimport type { MIConfig } from \"./types\";\n\n/**\n * Compute Resource Allocation index between neighbourhoods of two nodes.\n *\n * @param graph - Source graph\n * @param source - Source node ID\n * @param target - Target node ID\n * @param config - Optional configuration\n * @returns Resource Allocation index (normalised to [0, 1] if configured)\n */\nexport function resourceAllocation<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tsource: NodeId,\n\ttarget: NodeId,\n\tconfig?: MIConfig,\n): number {\n\tconst { epsilon = 1e-10, normalise = true } = config ?? {};\n\n\t// Get neighbourhoods, excluding opposite endpoint\n\tconst sourceNeighbours = neighbourSet(graph, source, target);\n\tconst targetNeighbours = neighbourSet(graph, target, source);\n\n\t// Compute common neighbours\n\tconst commonNeighbours = neighbourIntersection(\n\t\tsourceNeighbours,\n\t\ttargetNeighbours,\n\t);\n\n\t// Sum inverse degrees of common neighbours\n\tlet score = 0;\n\tfor (const neighbour of commonNeighbours) {\n\t\tconst degree = graph.degree(neighbour);\n\t\tif (degree > 0) {\n\t\t\tscore += 1 / degree;\n\t\t}\n\t}\n\n\t// Normalise to [0, 1] if requested\n\tif (normalise && commonNeighbours.size > 0) {\n\t\t// Max possible is when all common neighbours have minimum degree (1)\n\t\tconst maxScore = commonNeighbours.size;\n\t\tscore = score / maxScore;\n\t}\n\n\t// Apply epsilon floor for numerical stability\n\treturn Math.max(epsilon, score);\n}\n\n/**\n * Async variant of Resource Allocation index for use with async graph data sources.\n *\n * Fetches both neighbourhoods concurrently, then fetches degree for each common\n * neighbour to compute the inverse-degree weighted sum.\n */\nexport async function resourceAllocationAsync<\n\tN extends NodeData,\n\tE extends EdgeData,\n>(\n\tgraph: AsyncReadableGraph<N, E>,\n\tsource: NodeId,\n\ttarget: NodeId,\n\tconfig?: MIConfig,\n): Promise<number> {\n\tconst { epsilon = 1e-10, normalise = true } = config ?? {};\n\n\t// Fetch both neighbourhoods in parallel\n\tconst [sourceArr, targetArr] = await Promise.all([\n\t\tcollectAsyncIterable(graph.neighbours(source)),\n\t\tcollectAsyncIterable(graph.neighbours(target)),\n\t]);\n\n\tconst srcSet = new Set(sourceArr.filter((n) => n !== target));\n\tconst tgtSet = new Set(targetArr.filter((n) => n !== source));\n\n\t// Find common neighbours\n\tconst commonNeighbours: NodeId[] = [];\n\tfor (const n of srcSet) {\n\t\tif (tgtSet.has(n)) commonNeighbours.push(n);\n\t}\n\n\tif (commonNeighbours.length === 0) {\n\t\treturn epsilon;\n\t}\n\n\t// Fetch degrees of all common neighbours in parallel\n\tconst degrees = await Promise.all(\n\t\tcommonNeighbours.map((n) => graph.degree(n)),\n\t);\n\n\t// Sum inverse degrees\n\tlet score = 0;\n\tfor (const degree of degrees) {\n\t\tif (degree > 0) {\n\t\t\tscore += 1 / degree;\n\t\t}\n\t}\n\n\t// Normalise to [0, 1] if requested\n\tif (normalise) {\n\t\t// Max possible is when all common neighbours have minimum degree (1)\n\t\tconst maxScore = commonNeighbours.length;\n\t\tscore = score / maxScore;\n\t}\n\n\treturn Math.max(epsilon, score);\n}\n","/**\n * Overlap Coefficient for edge salience.\n *\n * Minimum-based neighbourhood overlap:\n * MI(u,v) = |N(u) ∩ N(v)| / min(|N(u)|, |N(v)|)\n *\n * Range: [0, 1]\n * - 0: No shared neighbours or one node has no neighbours\n * - 1: One neighbourhood is a subset of the other\n */\n\nimport type { NodeId, NodeData, EdgeData, ReadableGraph } from \"../../graph\";\nimport type { AsyncReadableGraph } from \"../../graph/async-interfaces\";\nimport { neighbourSet, neighbourOverlap } from \"../../utils\";\nimport { collectAsyncIterable } from \"../../async/utils\";\nimport type { MIConfig } from \"./types\";\n\n/**\n * Compute Overlap Coefficient between neighbourhoods of two nodes.\n *\n * @param graph - Source graph\n * @param source - Source node ID\n * @param target - Target node ID\n * @param config - Optional configuration\n * @returns Overlap Coefficient in [0, 1]\n */\nexport function overlapCoefficient<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tsource: NodeId,\n\ttarget: NodeId,\n\tconfig?: MIConfig,\n): number {\n\tconst { epsilon = 1e-10 } = config ?? {};\n\n\t// Get neighbourhoods, excluding opposite endpoint\n\tconst sourceNeighbours = neighbourSet(graph, source, target);\n\tconst targetNeighbours = neighbourSet(graph, target, source);\n\n\t// Compute intersection size\n\tconst { intersection } = neighbourOverlap(sourceNeighbours, targetNeighbours);\n\n\t// Compute denominator: min(|N(u)|, |N(v)|)\n\tconst denominator = Math.min(sourceNeighbours.size, targetNeighbours.size);\n\n\t// Avoid division by zero\n\tif (denominator === 0) {\n\t\treturn 0;\n\t}\n\n\tconst score = intersection / denominator;\n\n\t// Apply epsilon floor for numerical stability\n\treturn Math.max(epsilon, score);\n}\n\n/**\n * Async variant of Overlap Coefficient for use with async graph data sources.\n *\n * Fetches both neighbourhoods concurrently, then applies the same formula.\n */\nexport async function overlapCoefficientAsync<\n\tN extends NodeData,\n\tE extends EdgeData,\n>(\n\tgraph: AsyncReadableGraph<N, E>,\n\tsource: NodeId,\n\ttarget: NodeId,\n\tconfig?: MIConfig,\n): Promise<number> {\n\tconst { epsilon = 1e-10 } = config ?? {};\n\n\t// Fetch both neighbourhoods in parallel\n\tconst [sourceArr, targetArr] = await Promise.all([\n\t\tcollectAsyncIterable(graph.neighbours(source)),\n\t\tcollectAsyncIterable(graph.neighbours(target)),\n\t]);\n\n\tconst srcSet = new Set(sourceArr.filter((n) => n !== target));\n\tconst tgtSet = new Set(targetArr.filter((n) => n !== source));\n\n\t// Compute intersection size\n\tlet intersection = 0;\n\tfor (const n of srcSet) {\n\t\tif (tgtSet.has(n)) intersection++;\n\t}\n\n\t// Compute denominator: min(|N(u)|, |N(v)|)\n\tconst denominator = Math.min(srcSet.size, tgtSet.size);\n\n\t// Avoid division by zero\n\tif (denominator === 0) {\n\t\treturn 0;\n\t}\n\n\tconst score = intersection / denominator;\n\n\treturn Math.max(epsilon, score);\n}\n","/**\n * Hub Promoted index for edge salience.\n *\n * Hub-promoting neighbourhood overlap:\n * MI(u,v) = |N(u) ∩ N(v)| / min(deg(u), deg(v))\n *\n * Range: [0, 1]\n * Uses node degrees rather than neighbourhood set sizes.\n * Similar to Overlap Coefficient but normalised by degree.\n */\n\nimport type { NodeId, NodeData, EdgeData, ReadableGraph } from \"../../graph\";\nimport type { AsyncReadableGraph } from \"../../graph/async-interfaces\";\nimport { neighbourSet, neighbourOverlap } from \"../../utils\";\nimport { collectAsyncIterable } from \"../../async/utils\";\nimport type { MIConfig } from \"./types\";\n\n/**\n * Compute Hub Promoted index between neighbourhoods of two nodes.\n *\n * @param graph - Source graph\n * @param source - Source node ID\n * @param target - Target node ID\n * @param config - Optional configuration\n * @returns Hub Promoted index in [0, 1]\n */\nexport function hubPromoted<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tsource: NodeId,\n\ttarget: NodeId,\n\tconfig?: MIConfig,\n): number {\n\tconst { epsilon = 1e-10 } = config ?? {};\n\n\t// Get neighbourhoods, excluding opposite endpoint\n\tconst sourceNeighbours = neighbourSet(graph, source, target);\n\tconst targetNeighbours = neighbourSet(graph, target, source);\n\n\t// Compute intersection size\n\tconst { intersection } = neighbourOverlap(sourceNeighbours, targetNeighbours);\n\n\t// Compute denominator using actual degrees\n\tconst sourceDegree = graph.degree(source);\n\tconst targetDegree = graph.degree(target);\n\tconst denominator = Math.min(sourceDegree, targetDegree);\n\n\t// Avoid division by zero\n\tif (denominator === 0) {\n\t\treturn 0;\n\t}\n\n\tconst score = intersection / denominator;\n\n\t// Apply epsilon floor for numerical stability\n\treturn Math.max(epsilon, score);\n}\n\n/**\n * Async variant of Hub Promoted index for use with async graph data sources.\n *\n * Fetches both neighbourhoods and degrees concurrently, then applies the same formula.\n */\nexport async function hubPromotedAsync<N extends NodeData, E extends EdgeData>(\n\tgraph: AsyncReadableGraph<N, E>,\n\tsource: NodeId,\n\ttarget: NodeId,\n\tconfig?: MIConfig,\n): Promise<number> {\n\tconst { epsilon = 1e-10 } = config ?? {};\n\n\t// Fetch both neighbourhoods and degrees in parallel\n\tconst [sourceArr, targetArr, sourceDegree, targetDegree] = await Promise.all([\n\t\tcollectAsyncIterable(graph.neighbours(source)),\n\t\tcollectAsyncIterable(graph.neighbours(target)),\n\t\tgraph.degree(source),\n\t\tgraph.degree(target),\n\t]);\n\n\tconst srcSet = new Set(sourceArr.filter((n) => n !== target));\n\tconst tgtSet = new Set(targetArr.filter((n) => n !== source));\n\n\t// Compute intersection size\n\tlet intersection = 0;\n\tfor (const n of srcSet) {\n\t\tif (tgtSet.has(n)) intersection++;\n\t}\n\n\t// Compute denominator using actual degrees\n\tconst denominator = Math.min(sourceDegree, targetDegree);\n\n\t// Avoid division by zero\n\tif (denominator === 0) {\n\t\treturn 0;\n\t}\n\n\tconst score = intersection / denominator;\n\n\treturn Math.max(epsilon, score);\n}\n","/**\n * SCALE (Structural Coherence via Adjacency Lattice Entropy) MI variant.\n *\n * Density-normalised Jaccard, correcting for graph density variation.\n * Formula: MI(u,v) = Jaccard(u,v) / ρ(G)\n *\n * where ρ(G) = 2 * |E| / (|V| * (|V| - 1)) for undirected graphs\n *       ρ(G) = |E| / (|V| * (|V| - 1)) for directed graphs\n *\n * Range: [0, ∞) but typically scales with graph density\n */\n\nimport type { NodeId, NodeData, EdgeData, ReadableGraph } from \"../../graph\";\nimport type { AsyncReadableGraph } from \"../../graph/async-interfaces\";\nimport { computeJaccard } from \"../../utils\";\nimport { collectAsyncIterable } from \"../../async/utils\";\nimport type { MIConfig } from \"./types\";\n\n/**\n * Compute SCALE MI between two nodes.\n */\nexport function scale<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tsource: NodeId,\n\ttarget: NodeId,\n\tconfig?: MIConfig,\n): number {\n\tconst { epsilon = 1e-10 } = config ?? {};\n\n\tconst { jaccard: jaccardScore } = computeJaccard(graph, source, target);\n\n\t// Compute graph density\n\tconst n = graph.nodeCount;\n\tconst m = graph.edgeCount;\n\n\t// ρ(G) = 2|E| / (|V|(|V|-1)) for undirected; |E| / (|V|(|V|-1)) for directed\n\tconst possibleEdges = n * (n - 1);\n\tconst density =\n\t\tpossibleEdges > 0 ? (graph.directed ? m : 2 * m) / possibleEdges : 0;\n\n\t// Avoid division by zero: if density is 0, fall back to epsilon\n\tif (density === 0) {\n\t\treturn epsilon;\n\t}\n\n\tconst score = jaccardScore / density;\n\n\t// Apply epsilon floor for numerical stability\n\treturn Math.max(epsilon, score);\n}\n\n/**\n * Async variant of SCALE MI for use with async graph data sources.\n *\n * Fetches both neighbourhoods, node count, and edge count concurrently.\n */\nexport async function scaleAsync<N extends NodeData, E extends EdgeData>(\n\tgraph: AsyncReadableGraph<N, E>,\n\tsource: NodeId,\n\ttarget: NodeId,\n\tconfig?: MIConfig,\n): Promise<number> {\n\tconst { epsilon = 1e-10 } = config ?? {};\n\n\t// Fetch neighbourhoods, node count, and edge count in parallel\n\tconst [sourceArr, targetArr, n, m] = await Promise.all([\n\t\tcollectAsyncIterable(graph.neighbours(source)),\n\t\tcollectAsyncIterable(graph.neighbours(target)),\n\t\tgraph.nodeCount,\n\t\tgraph.edgeCount,\n\t]);\n\n\tconst srcSet = new Set(sourceArr.filter((node) => node !== target));\n\tconst tgtSet = new Set(targetArr.filter((node) => node !== source));\n\n\t// Compute Jaccard from sets\n\tlet intersection = 0;\n\tfor (const node of srcSet) {\n\t\tif (tgtSet.has(node)) intersection++;\n\t}\n\tconst union = srcSet.size + tgtSet.size - intersection;\n\tconst jaccardScore = union > 0 ? intersection / union : 0;\n\n\t// Compute graph density\n\t// ρ(G) = 2|E| / (|V|(|V|-1)) for undirected; |E| / (|V|(|V|-1)) for directed\n\tconst possibleEdges = n * (n - 1);\n\tconst density =\n\t\tpossibleEdges > 0 ? (graph.directed ? m : 2 * m) / possibleEdges : 0;\n\n\t// Avoid division by zero: if density is 0, fall back to epsilon\n\tif (density === 0) {\n\t\treturn epsilon;\n\t}\n\n\tconst score = jaccardScore / density;\n\n\treturn Math.max(epsilon, score);\n}\n","/**\n * SKEW (Structural Kernel Entropy Weighting) MI variant.\n *\n * IDF-style rarity weighting on endpoints, applied to Jaccard base.\n * Formula: MI(u,v) = Jaccard(u,v) * log(N/deg(u)+1) * log(N/deg(v)+1)\n *\n * Range: [0, ∞) but typically [0, 1] for well-connected graphs\n */\n\nimport type { NodeId, NodeData, EdgeData, ReadableGraph } from \"../../graph\";\nimport type { AsyncReadableGraph } from \"../../graph/async-interfaces\";\nimport { computeJaccard } from \"../../utils\";\nimport { collectAsyncIterable } from \"../../async/utils\";\nimport type { MIConfig } from \"./types\";\n\n/**\n * Compute SKEW MI between two nodes.\n */\nexport function skew<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tsource: NodeId,\n\ttarget: NodeId,\n\tconfig?: MIConfig,\n): number {\n\tconst { epsilon = 1e-10 } = config ?? {};\n\n\tconst { jaccard: jaccardScore } = computeJaccard(graph, source, target);\n\n\t// Compute IDF-style weights for endpoints\n\tconst N = graph.nodeCount;\n\tconst sourceDegree = graph.degree(source);\n\tconst targetDegree = graph.degree(target);\n\n\tconst sourceIdf = Math.log(N / (sourceDegree + 1));\n\tconst targetIdf = Math.log(N / (targetDegree + 1));\n\n\tconst score = jaccardScore * sourceIdf * targetIdf;\n\n\t// Apply epsilon floor for numerical stability\n\treturn Math.max(epsilon, score);\n}\n\n/**\n * Async variant of SKEW MI for use with async graph data sources.\n *\n * Fetches both neighbourhoods, degrees, and node count concurrently.\n */\nexport async function skewAsync<N extends NodeData, E extends EdgeData>(\n\tgraph: AsyncReadableGraph<N, E>,\n\tsource: NodeId,\n\ttarget: NodeId,\n\tconfig?: MIConfig,\n): Promise<number> {\n\tconst { epsilon = 1e-10 } = config ?? {};\n\n\t// Fetch neighbourhoods, node count, and endpoint degrees in parallel\n\tconst [sourceArr, targetArr, N, sourceDegree, targetDegree] =\n\t\tawait Promise.all([\n\t\t\tcollectAsyncIterable(graph.neighbours(source)),\n\t\t\tcollectAsyncIterable(graph.neighbours(target)),\n\t\t\tgraph.nodeCount,\n\t\t\tgraph.degree(source),\n\t\t\tgraph.degree(target),\n\t\t]);\n\n\tconst srcSet = new Set(sourceArr.filter((n) => n !== target));\n\tconst tgtSet = new Set(targetArr.filter((n) => n !== source));\n\n\t// Compute Jaccard from sets\n\tlet intersection = 0;\n\tfor (const n of srcSet) {\n\t\tif (tgtSet.has(n)) intersection++;\n\t}\n\tconst union = srcSet.size + tgtSet.size - intersection;\n\tconst jaccardScore = union > 0 ? intersection / union : 0;\n\n\t// Compute IDF-style weights for endpoints\n\tconst sourceIdf = Math.log(N / (sourceDegree + 1));\n\tconst targetIdf = Math.log(N / (targetDegree + 1));\n\n\tconst score = jaccardScore * sourceIdf * targetIdf;\n\n\treturn Math.max(epsilon, score);\n}\n","/**\n * SPAN (Structural Pattern ANalysis) MI variant.\n *\n * Clustering-coefficient penalty, favouring bridge edges.\n * Formula: MI(u,v) = Jaccard(u,v) * (1 - max(cc(u), cc(v)))\n *\n * Nodes with high clustering coefficient are tightly embedded in triangles;\n * edges between such nodes are less likely to be bridge edges. This variant\n * downweights such edges, favouring paths through bridge edges.\n *\n * Range: [0, 1]\n */\n\nimport type { NodeId, NodeData, EdgeData, ReadableGraph } from \"../../graph\";\nimport type { AsyncReadableGraph } from \"../../graph/async-interfaces\";\nimport { computeJaccard } from \"../../utils\";\nimport { localClusteringCoefficient } from \"../../utils\";\nimport { collectAsyncIterable } from \"../../async/utils\";\nimport type { MIConfig } from \"./types\";\n\n/**\n * Compute SPAN MI between two nodes.\n */\nexport function span<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tsource: NodeId,\n\ttarget: NodeId,\n\tconfig?: MIConfig,\n): number {\n\tconst { epsilon = 1e-10 } = config ?? {};\n\n\tconst { jaccard: jaccardScore } = computeJaccard(graph, source, target);\n\n\t// Compute clustering coefficients\n\tconst sourceCc = localClusteringCoefficient(graph, source);\n\tconst targetCc = localClusteringCoefficient(graph, target);\n\n\t// Apply bridge penalty: downweight edges between highly-embedded nodes\n\tconst bridgePenalty = 1 - Math.max(sourceCc, targetCc);\n\n\tconst score = jaccardScore * bridgePenalty;\n\n\t// Apply epsilon floor for numerical stability\n\treturn Math.max(epsilon, score);\n}\n\n/**\n * Async variant of SPAN MI for use with async graph data sources.\n *\n * Fetches both neighbourhoods concurrently, then computes the clustering\n * coefficient for each endpoint from the collected neighbour arrays.\n */\nexport async function spanAsync<N extends NodeData, E extends EdgeData>(\n\tgraph: AsyncReadableGraph<N, E>,\n\tsource: NodeId,\n\ttarget: NodeId,\n\tconfig?: MIConfig,\n): Promise<number> {\n\tconst { epsilon = 1e-10 } = config ?? {};\n\n\t// Fetch both neighbourhoods in parallel\n\tconst [sourceArr, targetArr] = await Promise.all([\n\t\tcollectAsyncIterable(graph.neighbours(source)),\n\t\tcollectAsyncIterable(graph.neighbours(target)),\n\t]);\n\n\tconst srcSet = new Set(sourceArr.filter((n) => n !== target));\n\tconst tgtSet = new Set(targetArr.filter((n) => n !== source));\n\n\t// Compute Jaccard from sets\n\tlet intersection = 0;\n\tfor (const n of srcSet) {\n\t\tif (tgtSet.has(n)) intersection++;\n\t}\n\tconst union = srcSet.size + tgtSet.size - intersection;\n\tconst jaccardScore = union > 0 ? intersection / union : 0;\n\n\t// Compute clustering coefficients for source and target asynchronously.\n\t// CC(v) = 2 * |triangles| / (deg * (deg - 1))\n\t// We need to check which pairs of neighbours are connected.\n\tconst computeClusteringCoefficient = async (\n\t\tnodeId: NodeId,\n\t\tneighbourArr: readonly NodeId[],\n\t): Promise<number> => {\n\t\tconst degree = neighbourArr.length;\n\t\tif (degree < 2) return 0;\n\n\t\t// Check each pair of neighbours for connectivity in parallel\n\t\tconst pairs: [NodeId, NodeId][] = [];\n\t\tfor (let i = 0; i < neighbourArr.length; i++) {\n\t\t\tfor (let j = i + 1; j < neighbourArr.length; j++) {\n\t\t\t\tconst u = neighbourArr[i];\n\t\t\t\tconst v = neighbourArr[j];\n\t\t\t\tif (u !== undefined && v !== undefined) {\n\t\t\t\t\tpairs.push([u, v]);\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\n\t\tconst edgeResults = await Promise.all(\n\t\t\tpairs.flatMap(([u, v]) => [graph.getEdge(u, v), graph.getEdge(v, u)]),\n\t\t);\n\n\t\tlet triangleCount = 0;\n\t\tfor (let i = 0; i < pairs.length; i++) {\n\t\t\t// Each pair produced two edge lookups at indices 2*i and 2*i+1\n\t\t\tif (\n\t\t\t\tedgeResults[2 * i] !== undefined ||\n\t\t\t\tedgeResults[2 * i + 1] !== undefined\n\t\t\t) {\n\t\t\t\ttriangleCount++;\n\t\t\t}\n\t\t}\n\n\t\tconst possibleTriangles = (degree * (degree - 1)) / 2;\n\t\treturn triangleCount / possibleTriangles;\n\t};\n\n\tconst [sourceCc, targetCc] = await Promise.all([\n\t\tcomputeClusteringCoefficient(source, sourceArr),\n\t\tcomputeClusteringCoefficient(target, targetArr),\n\t]);\n\n\t// Apply bridge penalty: downweight edges between highly-embedded nodes\n\tconst bridgePenalty = 1 - Math.max(sourceCc, targetCc);\n\n\tconst score = jaccardScore * bridgePenalty;\n\n\treturn Math.max(epsilon, score);\n}\n","/**\n * ETCH (Edge Topology Coherence via Homophily) MI variant.\n *\n * Edge-type rarity weighting applied to Jaccard base.\n * Formula: MI(u,v) = Jaccard(u,v) * rarity(edgeType(u,v))\n * where rarity(t) = log(|E| / count(edges with type t))\n *\n * Edges of rare types (fewer instances in the graph) receive higher salience,\n * making discoveries across unusual edge relationships more significant.\n *\n * Range: [0, ∞) but typically [0, 1] for well-typed edges\n */\n\nimport type { NodeId, NodeData, EdgeData, ReadableGraph } from \"../../graph\";\nimport type { AsyncReadableGraph } from \"../../graph/async-interfaces\";\nimport { computeJaccard, countEdgesOfType } from \"../../utils\";\nimport { collectAsyncIterable } from \"../../async/utils\";\nimport type { MIConfig } from \"./types\";\n\n/**\n * Compute ETCH MI between two nodes.\n */\nexport function etch<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tsource: NodeId,\n\ttarget: NodeId,\n\tconfig?: MIConfig,\n): number {\n\tconst { epsilon = 1e-10 } = config ?? {};\n\n\tconst { jaccard: jaccardScore } = computeJaccard(graph, source, target);\n\n\t// Get edge between source and target\n\tconst edge = graph.getEdge(source, target);\n\n\t// If edge has no type or doesn't exist, fall back to Jaccard\n\tif (edge?.type === undefined) {\n\t\treturn Math.max(epsilon, jaccardScore);\n\t}\n\n\t// Compute edge rarity: log(total edges / edges of this type)\n\tconst edgeTypeCount = countEdgesOfType(graph, edge.type);\n\n\t// Avoid division by zero\n\tif (edgeTypeCount === 0) {\n\t\treturn Math.max(epsilon, jaccardScore);\n\t}\n\n\tconst rarity = Math.log(graph.edgeCount / edgeTypeCount);\n\tconst score = jaccardScore * rarity;\n\n\t// Apply epsilon floor for numerical stability\n\treturn Math.max(epsilon, score);\n}\n\n/**\n * Async variant of ETCH MI for use with async graph data sources.\n *\n * Fetches both neighbourhoods and edge data concurrently, then counts\n * edges of the same type by iterating the async edge stream.\n */\nexport async function etchAsync<N extends NodeData, E extends EdgeData>(\n\tgraph: AsyncReadableGraph<N, E>,\n\tsource: NodeId,\n\ttarget: NodeId,\n\tconfig?: MIConfig,\n): Promise<number> {\n\tconst { epsilon = 1e-10 } = config ?? {};\n\n\t// Fetch neighbourhoods and the source→target edge concurrently\n\tconst [sourceArr, targetArr, edge] = await Promise.all([\n\t\tcollectAsyncIterable(graph.neighbours(source)),\n\t\tcollectAsyncIterable(graph.neighbours(target)),\n\t\tgraph.getEdge(source, target),\n\t]);\n\n\tconst srcSet = new Set(sourceArr.filter((n) => n !== target));\n\tconst tgtSet = new Set(targetArr.filter((n) => n !== source));\n\n\t// Compute Jaccard from sets\n\tlet intersection = 0;\n\tfor (const n of srcSet) {\n\t\tif (tgtSet.has(n)) intersection++;\n\t}\n\tconst union = srcSet.size + tgtSet.size - intersection;\n\tconst jaccardScore = union > 0 ? intersection / union : 0;\n\n\t// If edge has no type or doesn't exist, fall back to Jaccard\n\tif (edge?.type === undefined) {\n\t\treturn Math.max(epsilon, jaccardScore);\n\t}\n\n\tconst edgeType = edge.type;\n\n\t// Count edges of this type by iterating all edges\n\tlet edgeTypeCount = 0;\n\tlet totalEdges = 0;\n\tfor await (const e of graph.edges()) {\n\t\ttotalEdges++;\n\t\tif (e.type === edgeType) edgeTypeCount++;\n\t}\n\n\t// Avoid division by zero\n\tif (edgeTypeCount === 0) {\n\t\treturn Math.max(epsilon, jaccardScore);\n\t}\n\n\tconst rarity = Math.log(totalEdges / edgeTypeCount);\n\tconst score = jaccardScore * rarity;\n\n\treturn Math.max(epsilon, score);\n}\n","/**\n * NOTCH (Neighbourhood Overlap Topology Coherence via Homophily) MI variant.\n *\n * Node-type rarity weighting applied to Jaccard base.\n * Formula: MI(u,v) = Jaccard(u,v) * rarity(nodeType(u)) * rarity(nodeType(v))\n * where rarity(t) = log(|V| / count(nodes with type t))\n *\n * Paths connecting nodes of rare types (fewer instances in the graph) receive higher\n * salience, making discoveries involving unusual node types more significant.\n *\n * Range: [0, ∞) but typically [0, 1] for well-typed nodes\n */\n\nimport type { NodeId, NodeData, EdgeData, ReadableGraph } from \"../../graph\";\nimport type { AsyncReadableGraph } from \"../../graph/async-interfaces\";\nimport { computeJaccard, countNodesOfType } from \"../../utils\";\nimport { collectAsyncIterable } from \"../../async/utils\";\nimport type { MIConfig } from \"./types\";\n\n/**\n * Compute NOTCH MI between two nodes.\n */\nexport function notch<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tsource: NodeId,\n\ttarget: NodeId,\n\tconfig?: MIConfig,\n): number {\n\tconst { epsilon = 1e-10 } = config ?? {};\n\n\tconst { jaccard: jaccardScore } = computeJaccard(graph, source, target);\n\n\t// Get node data\n\tconst sourceNode = graph.getNode(source);\n\tconst targetNode = graph.getNode(target);\n\n\t// If either node lacks a type, fall back to Jaccard\n\tif (sourceNode?.type === undefined || targetNode?.type === undefined) {\n\t\treturn Math.max(epsilon, jaccardScore);\n\t}\n\n\t// Compute node rarity: log(total nodes / nodes of this type)\n\tconst sourceTypeCount = countNodesOfType(graph, sourceNode.type);\n\tconst targetTypeCount = countNodesOfType(graph, targetNode.type);\n\n\t// Avoid division by zero\n\tif (sourceTypeCount === 0 || targetTypeCount === 0) {\n\t\treturn Math.max(epsilon, jaccardScore);\n\t}\n\n\tconst sourceRarity = Math.log(graph.nodeCount / sourceTypeCount);\n\tconst targetRarity = Math.log(graph.nodeCount / targetTypeCount);\n\n\tconst score = jaccardScore * sourceRarity * targetRarity;\n\n\t// Apply epsilon floor for numerical stability\n\treturn Math.max(epsilon, score);\n}\n\n/**\n * Async variant of NOTCH MI for use with async graph data sources.\n *\n * Fetches both neighbourhoods and node data concurrently, then counts\n * nodes of each type by iterating the async node stream.\n */\nexport async function notchAsync<N extends NodeData, E extends EdgeData>(\n\tgraph: AsyncReadableGraph<N, E>,\n\tsource: NodeId,\n\ttarget: NodeId,\n\tconfig?: MIConfig,\n): Promise<number> {\n\tconst { epsilon = 1e-10 } = config ?? {};\n\n\t// Fetch neighbourhoods and node data concurrently\n\tconst [sourceArr, targetArr, sourceNode, targetNode] = await Promise.all([\n\t\tcollectAsyncIterable(graph.neighbours(source)),\n\t\tcollectAsyncIterable(graph.neighbours(target)),\n\t\tgraph.getNode(source),\n\t\tgraph.getNode(target),\n\t]);\n\n\tconst srcSet = new Set(sourceArr.filter((n) => n !== target));\n\tconst tgtSet = new Set(targetArr.filter((n) => n !== source));\n\n\t// Compute Jaccard from sets\n\tlet intersection = 0;\n\tfor (const n of srcSet) {\n\t\tif (tgtSet.has(n)) intersection++;\n\t}\n\tconst union = srcSet.size + tgtSet.size - intersection;\n\tconst jaccardScore = union > 0 ? intersection / union : 0;\n\n\t// If either node lacks a type, fall back to Jaccard\n\tif (sourceNode?.type === undefined || targetNode?.type === undefined) {\n\t\treturn Math.max(epsilon, jaccardScore);\n\t}\n\n\tconst sourceType = sourceNode.type;\n\tconst targetType = targetNode.type;\n\n\t// Count nodes of each type and total nodes by iterating all node IDs\n\tlet totalNodes = 0;\n\tlet sourceTypeCount = 0;\n\tlet targetTypeCount = 0;\n\tfor await (const nodeId of graph.nodeIds()) {\n\t\ttotalNodes++;\n\t\tconst node = await graph.getNode(nodeId);\n\t\tif (node?.type === sourceType) sourceTypeCount++;\n\t\tif (node?.type === targetType) targetTypeCount++;\n\t}\n\n\t// Avoid division by zero\n\tif (sourceTypeCount === 0 || targetTypeCount === 0) {\n\t\treturn Math.max(epsilon, jaccardScore);\n\t}\n\n\tconst sourceRarity = Math.log(totalNodes / sourceTypeCount);\n\tconst targetRarity = Math.log(totalNodes / targetTypeCount);\n\n\tconst score = jaccardScore * sourceRarity * targetRarity;\n\n\treturn Math.max(epsilon, score);\n}\n","/**\n * Unified Adaptive MI - combines multiple MI signals dynamically.\n *\n * Adapts to graph structure by weighting different MI components\n * based on structural properties.\n *\n * Three-component weighted sum:\n * - Structural: Jaccard neighbourhood overlap\n * - Degree: Adamic-Adar inverse-log-degree weighting\n * - Overlap: Overlap coefficient (intersection / min degree)\n *\n * Range: [0, 1] - higher values indicate stronger association\n */\n\nimport type { NodeId, NodeData, EdgeData, ReadableGraph } from \"../../graph\";\nimport type { AsyncReadableGraph } from \"../../graph/async-interfaces\";\nimport { computeJaccard, neighbourOverlap } from \"../../utils\";\nimport { collectAsyncIterable } from \"../../async/utils\";\nimport type { AdaptiveMIConfig } from \"./types\";\nimport { adamicAdar } from \"./adamic-adar\";\nimport { adamicAdarAsync } from \"./adamic-adar\";\n\n/**\n * Compute unified adaptive MI between two connected nodes.\n *\n * Combines structural, degree, and overlap signals with\n * configurable weighting.\n *\n * @param graph - Source graph\n * @param source - Source node ID\n * @param target - Target node ID\n * @param config - Optional configuration with component weights\n * @returns Adaptive MI score in [0, 1]\n */\nexport function adaptive<N extends NodeData, E extends EdgeData>(\n\tgraph: ReadableGraph<N, E>,\n\tsource: NodeId,\n\ttarget: NodeId,\n\tconfig?: AdaptiveMIConfig,\n): number {\n\tconst {\n\t\tepsilon = 1e-10,\n\t\tstructuralWeight = 0.4,\n\t\tdegreeWeight = 0.3,\n\t\toverlapWeight = 0.3,\n\t} = config ?? {};\n\n\t// Compute Jaccard and retrieve neighbourhood sets for the overlap coefficient\n\tconst {\n\t\tjaccard: jaccardScore,\n\t\tsourceNeighbours,\n\t\ttargetNeighbours,\n\t} = computeJaccard(graph, source, target);\n\n\t// Component 1: Structural similarity (Jaccard)\n\t// Returns 0 only when both neighbourhood sets are empty (union = 0); otherwise applies epsilon floor\n\tconst structural =\n\t\tsourceNeighbours.size === 0 && targetNeighbours.size === 0\n\t\t\t? 0\n\t\t\t: Math.max(epsilon, jaccardScore);\n\n\t// Component 2: Degree-weighted association (Adamic-Adar, normalised)\n\tconst degreeComponent = adamicAdar(graph, source, target, {\n\t\tepsilon,\n\t\tnormalise: true,\n\t});\n\n\t// Component 3: Overlap coefficient\n\tlet overlap: number;\n\tif (sourceNeighbours.size > 0 && targetNeighbours.size > 0) {\n\t\tconst { intersection } = neighbourOverlap(\n\t\t\tsourceNeighbours,\n\t\t\ttargetNeighbours,\n\t\t);\n\t\tconst minDegree = Math.min(sourceNeighbours.size, targetNeighbours.size);\n\t\toverlap = minDegree > 0 ? intersection / minDegree : epsilon;\n\t} else {\n\t\toverlap = epsilon;\n\t}\n\n\t// Normalise weights\n\tconst totalWeight = structuralWeight + degreeWeight + overlapWeight;\n\n\t// Weighted combination\n\tconst score =\n\t\t(structuralWeight * structural +\n\t\t\tdegreeWeight * degreeComponent +\n\t\t\toverlapWeight * overlap) /\n\t\ttotalWeight;\n\n\treturn Math.max(epsilon, Math.min(1, score));\n}\n\n/**\n * Async variant of Adaptive MI for use with async graph data sources.\n *\n * Fetches both neighbourhoods concurrently, then delegates degree-weighted\n * component to the async Adamic-Adar variant.\n */\nexport async function adaptiveAsync<N extends NodeData, E extends EdgeData>(\n\tgraph: AsyncReadableGraph<N, E>,\n\tsource: NodeId,\n\ttarget: NodeId,\n\tconfig?: AdaptiveMIConfig,\n): Promise<number> {\n\tconst {\n\t\tepsilon = 1e-10,\n\t\tstructuralWeight = 0.4,\n\t\tdegreeWeight = 0.3,\n\t\toverlapWeight = 0.3,\n\t} = config ?? {};\n\n\t// Fetch both neighbourhoods in parallel alongside the Adamic-Adar component\n\tconst [sourceArr, targetArr, degreeComponent] = await Promise.all([\n\t\tcollectAsyncIterable(graph.neighbours(source)),\n\t\tcollectAsyncIterable(graph.neighbours(target)),\n\t\tadamicAdarAsync(graph, source, target, { epsilon, normalise: true }),\n\t]);\n\n\tconst srcSet = new Set(sourceArr.filter((n) => n !== target));\n\tconst tgtSet = new Set(targetArr.filter((n) => n !== source));\n\n\t// Compute Jaccard from sets\n\tlet intersection = 0;\n\tfor (const n of srcSet) {\n\t\tif (tgtSet.has(n)) intersection++;\n\t}\n\tconst union = srcSet.size + tgtSet.size - intersection;\n\tconst jaccardScore = union > 0 ? intersection / union : 0;\n\n\t// Component 1: Structural similarity (Jaccard)\n\tconst structural =\n\t\tsrcSet.size === 0 && tgtSet.size === 0\n\t\t\t? 0\n\t\t\t: Math.max(epsilon, jaccardScore);\n\n\t// Component 3: Overlap coefficient\n\tlet overlap: number;\n\tif (srcSet.size > 0 && tgtSet.size > 0) {\n\t\tconst minDegree = Math.min(srcSet.size, tgtSet.size);\n\t\toverlap = minDegree > 0 ? intersection / minDegree : epsilon;\n\t} else {\n\t\toverlap = epsilon;\n\t}\n\n\t// Normalise weights\n\tconst totalWeight = structuralWeight + degreeWeight + overlapWeight;\n\n\t// Weighted combination\n\tconst score =\n\t\t(structuralWeight * structural +\n\t\t\tdegreeWeight * degreeComponent +\n\t\t\toverlapWeight * overlap) /\n\t\ttotalWeight;\n\n\treturn Math.max(epsilon, Math.min(1, score));\n}\n"],"mappings":";;;;;;;;;;;;;;AAyBA,SAAgB,WACf,OACA,QACA,QACA,QACS;CACT,MAAM,EAAE,UAAU,OAAO,YAAY,SAAS,UAAU,EAAE;CAO1D,MAAM,mBAAmB,gBAAA,sBAJA,gBAAA,aAAa,OAAO,QAAQ,OAAO,EACnC,gBAAA,aAAa,OAAO,QAAQ,OAAO,CAM3D;CAGD,IAAI,QAAQ;AACZ,MAAK,MAAM,aAAa,kBAAkB;EACzC,MAAM,SAAS,MAAM,OAAO,UAAU;AACtC,WAAS,IAAI,KAAK,IAAI,SAAS,EAAE;;AAIlC,KAAI,aAAa,iBAAiB,OAAO,GAAG;EAG3C,MAAM,WAAW,iBAAiB,OAAO,KAAK,IAAI,EAAE;AACpD,UAAQ,QAAQ;;AAIjB,QAAO,KAAK,IAAI,SAAS,MAAM;;;;;;;;AAShC,eAAsB,gBACrB,OACA,QACA,QACA,QACkB;CAClB,MAAM,EAAE,UAAU,OAAO,YAAY,SAAS,UAAU,EAAE;CAG1D,MAAM,CAAC,WAAW,aAAa,MAAM,QAAQ,IAAI,CAChD,cAAA,qBAAqB,MAAM,WAAW,OAAO,CAAC,EAC9C,cAAA,qBAAqB,MAAM,WAAW,OAAO,CAAC,CAC9C,CAAC;CAEF,MAAM,SAAS,IAAI,IAAI,UAAU,QAAQ,MAAM,MAAM,OAAO,CAAC;CAC7D,MAAM,SAAS,IAAI,IAAI,UAAU,QAAQ,MAAM,MAAM,OAAO,CAAC;CAG7D,MAAM,mBAA6B,EAAE;AACrC,MAAK,MAAM,KAAK,OACf,KAAI,OAAO,IAAI,EAAE,CAAE,kBAAiB,KAAK,EAAE;AAG5C,KAAI,iBAAiB,WAAW,EAC/B,QAAO;CAIR,MAAM,UAAU,MAAM,QAAQ,IAC7B,iBAAiB,KAAK,MAAM,MAAM,OAAO,EAAE,CAAC,CAC5C;CAGD,IAAI,QAAQ;AACZ,MAAK,MAAM,UAAU,QACpB,UAAS,IAAI,KAAK,IAAI,SAAS,EAAE;AAIlC,KAAI,WAAW;EAGd,MAAM,WAAW,iBAAiB,SAAS,KAAK,IAAI,EAAE;AACtD,UAAQ,QAAQ;;AAGjB,QAAO,KAAK,IAAI,SAAS,MAAM;;;;;;;;;;;;;ACxFhC,SAAgB,OACf,OACA,QACA,QACA,QACS;CACT,MAAM,EAAE,UAAU,UAAU,UAAU,EAAE;CAGxC,MAAM,mBAAmB,gBAAA,aAAa,OAAO,QAAQ,OAAO;CAC5D,MAAM,mBAAmB,gBAAA,aAAa,OAAO,QAAQ,OAAO;CAG5D,MAAM,EAAE,iBAAiB,gBAAA,iBAAiB,kBAAkB,iBAAiB;CAG7E,MAAM,cACL,KAAK,KAAK,iBAAiB,KAAK,GAAG,KAAK,KAAK,iBAAiB,KAAK;AAGpE,KAAI,gBAAgB,EACnB,QAAO;CAGR,MAAM,QAAQ,eAAe;AAG7B,QAAO,KAAK,IAAI,SAAS,MAAM;;;;;;;AAQhC,eAAsB,YACrB,OACA,QACA,QACA,QACkB;CAClB,MAAM,EAAE,UAAU,UAAU,UAAU,EAAE;CAGxC,MAAM,CAAC,WAAW,aAAa,MAAM,QAAQ,IAAI,CAChD,cAAA,qBAAqB,MAAM,WAAW,OAAO,CAAC,EAC9C,cAAA,qBAAqB,MAAM,WAAW,OAAO,CAAC,CAC9C,CAAC;CAEF,MAAM,SAAS,IAAI,IAAI,UAAU,QAAQ,MAAM,MAAM,OAAO,CAAC;CAC7D,MAAM,SAAS,IAAI,IAAI,UAAU,QAAQ,MAAM,MAAM,OAAO,CAAC;CAG7D,IAAI,eAAe;AACnB,MAAK,MAAM,KAAK,OACf,KAAI,OAAO,IAAI,EAAE,CAAE;CAIpB,MAAM,cAAc,KAAK,KAAK,OAAO,KAAK,GAAG,KAAK,KAAK,OAAO,KAAK;AAGnE,KAAI,gBAAgB,EACnB,QAAO;CAGR,MAAM,QAAQ,eAAe;AAE7B,QAAO,KAAK,IAAI,SAAS,MAAM;;;;;;;;;;;;;ACpEhC,SAAgB,SACf,OACA,QACA,QACA,QACS;CACT,MAAM,EAAE,UAAU,UAAU,UAAU,EAAE;CAGxC,MAAM,mBAAmB,gBAAA,aAAa,OAAO,QAAQ,OAAO;CAC5D,MAAM,mBAAmB,gBAAA,aAAa,OAAO,QAAQ,OAAO;CAG5D,MAAM,EAAE,iBAAiB,gBAAA,iBAAiB,kBAAkB,iBAAiB;CAG7E,MAAM,cAAc,iBAAiB,OAAO,iBAAiB;AAG7D,KAAI,gBAAgB,EACnB,QAAO;CAGR,MAAM,QAAS,IAAI,eAAgB;AAGnC,QAAO,KAAK,IAAI,SAAS,MAAM;;;;;;;AAQhC,eAAsB,cACrB,OACA,QACA,QACA,QACkB;CAClB,MAAM,EAAE,UAAU,UAAU,UAAU,EAAE;CAGxC,MAAM,CAAC,WAAW,aAAa,MAAM,QAAQ,IAAI,CAChD,cAAA,qBAAqB,MAAM,WAAW,OAAO,CAAC,EAC9C,cAAA,qBAAqB,MAAM,WAAW,OAAO,CAAC,CAC9C,CAAC;CAEF,MAAM,SAAS,IAAI,IAAI,UAAU,QAAQ,MAAM,MAAM,OAAO,CAAC;CAC7D,MAAM,SAAS,IAAI,IAAI,UAAU,QAAQ,MAAM,MAAM,OAAO,CAAC;CAG7D,IAAI,eAAe;AACnB,MAAK,MAAM,KAAK,OACf,KAAI,OAAO,IAAI,EAAE,CAAE;CAIpB,MAAM,cAAc,OAAO,OAAO,OAAO;AAGzC,KAAI,gBAAgB,EACnB,QAAO;CAGR,MAAM,QAAS,IAAI,eAAgB;AAEnC,QAAO,KAAK,IAAI,SAAS,MAAM;;;;;;;;;;;;;ACpEhC,SAAgB,mBACf,OACA,QACA,QACA,QACS;CACT,MAAM,EAAE,UAAU,OAAO,YAAY,SAAS,UAAU,EAAE;CAO1D,MAAM,mBAAmB,gBAAA,sBAJA,gBAAA,aAAa,OAAO,QAAQ,OAAO,EACnC,gBAAA,aAAa,OAAO,QAAQ,OAAO,CAM3D;CAGD,IAAI,QAAQ;AACZ,MAAK,MAAM,aAAa,kBAAkB;EACzC,MAAM,SAAS,MAAM,OAAO,UAAU;AACtC,MAAI,SAAS,EACZ,UAAS,IAAI;;AAKf,KAAI,aAAa,iBAAiB,OAAO,GAAG;EAE3C,MAAM,WAAW,iBAAiB;AAClC,UAAQ,QAAQ;;AAIjB,QAAO,KAAK,IAAI,SAAS,MAAM;;;;;;;;AAShC,eAAsB,wBAIrB,OACA,QACA,QACA,QACkB;CAClB,MAAM,EAAE,UAAU,OAAO,YAAY,SAAS,UAAU,EAAE;CAG1D,MAAM,CAAC,WAAW,aAAa,MAAM,QAAQ,IAAI,CAChD,cAAA,qBAAqB,MAAM,WAAW,OAAO,CAAC,EAC9C,cAAA,qBAAqB,MAAM,WAAW,OAAO,CAAC,CAC9C,CAAC;CAEF,MAAM,SAAS,IAAI,IAAI,UAAU,QAAQ,MAAM,MAAM,OAAO,CAAC;CAC7D,MAAM,SAAS,IAAI,IAAI,UAAU,QAAQ,MAAM,MAAM,OAAO,CAAC;CAG7D,MAAM,mBAA6B,EAAE;AACrC,MAAK,MAAM,KAAK,OACf,KAAI,OAAO,IAAI,EAAE,CAAE,kBAAiB,KAAK,EAAE;AAG5C,KAAI,iBAAiB,WAAW,EAC/B,QAAO;CAIR,MAAM,UAAU,MAAM,QAAQ,IAC7B,iBAAiB,KAAK,MAAM,MAAM,OAAO,EAAE,CAAC,CAC5C;CAGD,IAAI,QAAQ;AACZ,MAAK,MAAM,UAAU,QACpB,KAAI,SAAS,EACZ,UAAS,IAAI;AAKf,KAAI,WAAW;EAEd,MAAM,WAAW,iBAAiB;AAClC,UAAQ,QAAQ;;AAGjB,QAAO,KAAK,IAAI,SAAS,MAAM;;;;;;;;;;;;;AC7FhC,SAAgB,mBACf,OACA,QACA,QACA,QACS;CACT,MAAM,EAAE,UAAU,UAAU,UAAU,EAAE;CAGxC,MAAM,mBAAmB,gBAAA,aAAa,OAAO,QAAQ,OAAO;CAC5D,MAAM,mBAAmB,gBAAA,aAAa,OAAO,QAAQ,OAAO;CAG5D,MAAM,EAAE,iBAAiB,gBAAA,iBAAiB,kBAAkB,iBAAiB;CAG7E,MAAM,cAAc,KAAK,IAAI,iBAAiB,MAAM,iBAAiB,KAAK;AAG1E,KAAI,gBAAgB,EACnB,QAAO;CAGR,MAAM,QAAQ,eAAe;AAG7B,QAAO,KAAK,IAAI,SAAS,MAAM;;;;;;;AAQhC,eAAsB,wBAIrB,OACA,QACA,QACA,QACkB;CAClB,MAAM,EAAE,UAAU,UAAU,UAAU,EAAE;CAGxC,MAAM,CAAC,WAAW,aAAa,MAAM,QAAQ,IAAI,CAChD,cAAA,qBAAqB,MAAM,WAAW,OAAO,CAAC,EAC9C,cAAA,qBAAqB,MAAM,WAAW,OAAO,CAAC,CAC9C,CAAC;CAEF,MAAM,SAAS,IAAI,IAAI,UAAU,QAAQ,MAAM,MAAM,OAAO,CAAC;CAC7D,MAAM,SAAS,IAAI,IAAI,UAAU,QAAQ,MAAM,MAAM,OAAO,CAAC;CAG7D,IAAI,eAAe;AACnB,MAAK,MAAM,KAAK,OACf,KAAI,OAAO,IAAI,EAAE,CAAE;CAIpB,MAAM,cAAc,KAAK,IAAI,OAAO,MAAM,OAAO,KAAK;AAGtD,KAAI,gBAAgB,EACnB,QAAO;CAGR,MAAM,QAAQ,eAAe;AAE7B,QAAO,KAAK,IAAI,SAAS,MAAM;;;;;;;;;;;;;ACtEhC,SAAgB,YACf,OACA,QACA,QACA,QACS;CACT,MAAM,EAAE,UAAU,UAAU,UAAU,EAAE;CAOxC,MAAM,EAAE,iBAAiB,gBAAA,iBAJA,gBAAA,aAAa,OAAO,QAAQ,OAAO,EACnC,gBAAA,aAAa,OAAO,QAAQ,OAAO,CAGiB;CAG7E,MAAM,eAAe,MAAM,OAAO,OAAO;CACzC,MAAM,eAAe,MAAM,OAAO,OAAO;CACzC,MAAM,cAAc,KAAK,IAAI,cAAc,aAAa;AAGxD,KAAI,gBAAgB,EACnB,QAAO;CAGR,MAAM,QAAQ,eAAe;AAG7B,QAAO,KAAK,IAAI,SAAS,MAAM;;;;;;;AAQhC,eAAsB,iBACrB,OACA,QACA,QACA,QACkB;CAClB,MAAM,EAAE,UAAU,UAAU,UAAU,EAAE;CAGxC,MAAM,CAAC,WAAW,WAAW,cAAc,gBAAgB,MAAM,QAAQ,IAAI;EAC5E,cAAA,qBAAqB,MAAM,WAAW,OAAO,CAAC;EAC9C,cAAA,qBAAqB,MAAM,WAAW,OAAO,CAAC;EAC9C,MAAM,OAAO,OAAO;EACpB,MAAM,OAAO,OAAO;EACpB,CAAC;CAEF,MAAM,SAAS,IAAI,IAAI,UAAU,QAAQ,MAAM,MAAM,OAAO,CAAC;CAC7D,MAAM,SAAS,IAAI,IAAI,UAAU,QAAQ,MAAM,MAAM,OAAO,CAAC;CAG7D,IAAI,eAAe;AACnB,MAAK,MAAM,KAAK,OACf,KAAI,OAAO,IAAI,EAAE,CAAE;CAIpB,MAAM,cAAc,KAAK,IAAI,cAAc,aAAa;AAGxD,KAAI,gBAAgB,EACnB,QAAO;CAGR,MAAM,QAAQ,eAAe;AAE7B,QAAO,KAAK,IAAI,SAAS,MAAM;;;;;;;AC5EhC,SAAgB,MACf,OACA,QACA,QACA,QACS;CACT,MAAM,EAAE,UAAU,UAAU,UAAU,EAAE;CAExC,MAAM,EAAE,SAAS,iBAAiB,gBAAA,eAAe,OAAO,QAAQ,OAAO;CAGvE,MAAM,IAAI,MAAM;CAChB,MAAM,IAAI,MAAM;CAGhB,MAAM,gBAAgB,KAAK,IAAI;CAC/B,MAAM,UACL,gBAAgB,KAAK,MAAM,WAAW,IAAI,IAAI,KAAK,gBAAgB;AAGpE,KAAI,YAAY,EACf,QAAO;CAGR,MAAM,QAAQ,eAAe;AAG7B,QAAO,KAAK,IAAI,SAAS,MAAM;;;;;;;AAQhC,eAAsB,WACrB,OACA,QACA,QACA,QACkB;CAClB,MAAM,EAAE,UAAU,UAAU,UAAU,EAAE;CAGxC,MAAM,CAAC,WAAW,WAAW,GAAG,KAAK,MAAM,QAAQ,IAAI;EACtD,cAAA,qBAAqB,MAAM,WAAW,OAAO,CAAC;EAC9C,cAAA,qBAAqB,MAAM,WAAW,OAAO,CAAC;EAC9C,MAAM;EACN,MAAM;EACN,CAAC;CAEF,MAAM,SAAS,IAAI,IAAI,UAAU,QAAQ,SAAS,SAAS,OAAO,CAAC;CACnE,MAAM,SAAS,IAAI,IAAI,UAAU,QAAQ,SAAS,SAAS,OAAO,CAAC;CAGnE,IAAI,eAAe;AACnB,MAAK,MAAM,QAAQ,OAClB,KAAI,OAAO,IAAI,KAAK,CAAE;CAEvB,MAAM,QAAQ,OAAO,OAAO,OAAO,OAAO;CAC1C,MAAM,eAAe,QAAQ,IAAI,eAAe,QAAQ;CAIxD,MAAM,gBAAgB,KAAK,IAAI;CAC/B,MAAM,UACL,gBAAgB,KAAK,MAAM,WAAW,IAAI,IAAI,KAAK,gBAAgB;AAGpE,KAAI,YAAY,EACf,QAAO;CAGR,MAAM,QAAQ,eAAe;AAE7B,QAAO,KAAK,IAAI,SAAS,MAAM;;;;;;;AC9EhC,SAAgB,KACf,OACA,QACA,QACA,QACS;CACT,MAAM,EAAE,UAAU,UAAU,UAAU,EAAE;CAExC,MAAM,EAAE,SAAS,iBAAiB,gBAAA,eAAe,OAAO,QAAQ,OAAO;CAGvE,MAAM,IAAI,MAAM;CAChB,MAAM,eAAe,MAAM,OAAO,OAAO;CACzC,MAAM,eAAe,MAAM,OAAO,OAAO;CAEzC,MAAM,YAAY,KAAK,IAAI,KAAK,eAAe,GAAG;CAClD,MAAM,YAAY,KAAK,IAAI,KAAK,eAAe,GAAG;CAElD,MAAM,QAAQ,eAAe,YAAY;AAGzC,QAAO,KAAK,IAAI,SAAS,MAAM;;;;;;;AAQhC,eAAsB,UACrB,OACA,QACA,QACA,QACkB;CAClB,MAAM,EAAE,UAAU,UAAU,UAAU,EAAE;CAGxC,MAAM,CAAC,WAAW,WAAW,GAAG,cAAc,gBAC7C,MAAM,QAAQ,IAAI;EACjB,cAAA,qBAAqB,MAAM,WAAW,OAAO,CAAC;EAC9C,cAAA,qBAAqB,MAAM,WAAW,OAAO,CAAC;EAC9C,MAAM;EACN,MAAM,OAAO,OAAO;EACpB,MAAM,OAAO,OAAO;EACpB,CAAC;CAEH,MAAM,SAAS,IAAI,IAAI,UAAU,QAAQ,MAAM,MAAM,OAAO,CAAC;CAC7D,MAAM,SAAS,IAAI,IAAI,UAAU,QAAQ,MAAM,MAAM,OAAO,CAAC;CAG7D,IAAI,eAAe;AACnB,MAAK,MAAM,KAAK,OACf,KAAI,OAAO,IAAI,EAAE,CAAE;CAEpB,MAAM,QAAQ,OAAO,OAAO,OAAO,OAAO;CAC1C,MAAM,eAAe,QAAQ,IAAI,eAAe,QAAQ;CAGxD,MAAM,YAAY,KAAK,IAAI,KAAK,eAAe,GAAG;CAClD,MAAM,YAAY,KAAK,IAAI,KAAK,eAAe,GAAG;CAElD,MAAM,QAAQ,eAAe,YAAY;AAEzC,QAAO,KAAK,IAAI,SAAS,MAAM;;;;;;;AC3DhC,SAAgB,KACf,OACA,QACA,QACA,QACS;CACT,MAAM,EAAE,UAAU,UAAU,UAAU,EAAE;CAExC,MAAM,EAAE,SAAS,iBAAiB,gBAAA,eAAe,OAAO,QAAQ,OAAO;CAGvE,MAAM,WAAW,gBAAA,2BAA2B,OAAO,OAAO;CAC1D,MAAM,WAAW,gBAAA,2BAA2B,OAAO,OAAO;CAK1D,MAAM,QAAQ,gBAFQ,IAAI,KAAK,IAAI,UAAU,SAAS;AAKtD,QAAO,KAAK,IAAI,SAAS,MAAM;;;;;;;;AAShC,eAAsB,UACrB,OACA,QACA,QACA,QACkB;CAClB,MAAM,EAAE,UAAU,UAAU,UAAU,EAAE;CAGxC,MAAM,CAAC,WAAW,aAAa,MAAM,QAAQ,IAAI,CAChD,cAAA,qBAAqB,MAAM,WAAW,OAAO,CAAC,EAC9C,cAAA,qBAAqB,MAAM,WAAW,OAAO,CAAC,CAC9C,CAAC;CAEF,MAAM,SAAS,IAAI,IAAI,UAAU,QAAQ,MAAM,MAAM,OAAO,CAAC;CAC7D,MAAM,SAAS,IAAI,IAAI,UAAU,QAAQ,MAAM,MAAM,OAAO,CAAC;CAG7D,IAAI,eAAe;AACnB,MAAK,MAAM,KAAK,OACf,KAAI,OAAO,IAAI,EAAE,CAAE;CAEpB,MAAM,QAAQ,OAAO,OAAO,OAAO,OAAO;CAC1C,MAAM,eAAe,QAAQ,IAAI,eAAe,QAAQ;CAKxD,MAAM,+BAA+B,OACpC,QACA,iBACqB;EACrB,MAAM,SAAS,aAAa;AAC5B,MAAI,SAAS,EAAG,QAAO;EAGvB,MAAM,QAA4B,EAAE;AACpC,OAAK,IAAI,IAAI,GAAG,IAAI,aAAa,QAAQ,IACxC,MAAK,IAAI,IAAI,IAAI,GAAG,IAAI,aAAa,QAAQ,KAAK;GACjD,MAAM,IAAI,aAAa;GACvB,MAAM,IAAI,aAAa;AACvB,OAAI,MAAM,KAAA,KAAa,MAAM,KAAA,EAC5B,OAAM,KAAK,CAAC,GAAG,EAAE,CAAC;;EAKrB,MAAM,cAAc,MAAM,QAAQ,IACjC,MAAM,SAAS,CAAC,GAAG,OAAO,CAAC,MAAM,QAAQ,GAAG,EAAE,EAAE,MAAM,QAAQ,GAAG,EAAE,CAAC,CAAC,CACrE;EAED,IAAI,gBAAgB;AACpB,OAAK,IAAI,IAAI,GAAG,IAAI,MAAM,QAAQ,IAEjC,KACC,YAAY,IAAI,OAAO,KAAA,KACvB,YAAY,IAAI,IAAI,OAAO,KAAA,EAE3B;EAIF,MAAM,oBAAqB,UAAU,SAAS,KAAM;AACpD,SAAO,gBAAgB;;CAGxB,MAAM,CAAC,UAAU,YAAY,MAAM,QAAQ,IAAI,CAC9C,6BAA6B,QAAQ,UAAU,EAC/C,6BAA6B,QAAQ,UAAU,CAC/C,CAAC;CAKF,MAAM,QAAQ,gBAFQ,IAAI,KAAK,IAAI,UAAU,SAAS;AAItD,QAAO,KAAK,IAAI,SAAS,MAAM;;;;;;;AC1GhC,SAAgB,KACf,OACA,QACA,QACA,QACS;CACT,MAAM,EAAE,UAAU,UAAU,UAAU,EAAE;CAExC,MAAM,EAAE,SAAS,iBAAiB,gBAAA,eAAe,OAAO,QAAQ,OAAO;CAGvE,MAAM,OAAO,MAAM,QAAQ,QAAQ,OAAO;AAG1C,KAAI,MAAM,SAAS,KAAA,EAClB,QAAO,KAAK,IAAI,SAAS,aAAa;CAIvC,MAAM,gBAAgB,gBAAA,iBAAiB,OAAO,KAAK,KAAK;AAGxD,KAAI,kBAAkB,EACrB,QAAO,KAAK,IAAI,SAAS,aAAa;CAIvC,MAAM,QAAQ,eADC,KAAK,IAAI,MAAM,YAAY,cAAc;AAIxD,QAAO,KAAK,IAAI,SAAS,MAAM;;;;;;;;AAShC,eAAsB,UACrB,OACA,QACA,QACA,QACkB;CAClB,MAAM,EAAE,UAAU,UAAU,UAAU,EAAE;CAGxC,MAAM,CAAC,WAAW,WAAW,QAAQ,MAAM,QAAQ,IAAI;EACtD,cAAA,qBAAqB,MAAM,WAAW,OAAO,CAAC;EAC9C,cAAA,qBAAqB,MAAM,WAAW,OAAO,CAAC;EAC9C,MAAM,QAAQ,QAAQ,OAAO;EAC7B,CAAC;CAEF,MAAM,SAAS,IAAI,IAAI,UAAU,QAAQ,MAAM,MAAM,OAAO,CAAC;CAC7D,MAAM,SAAS,IAAI,IAAI,UAAU,QAAQ,MAAM,MAAM,OAAO,CAAC;CAG7D,IAAI,eAAe;AACnB,MAAK,MAAM,KAAK,OACf,KAAI,OAAO,IAAI,EAAE,CAAE;CAEpB,MAAM,QAAQ,OAAO,OAAO,OAAO,OAAO;CAC1C,MAAM,eAAe,QAAQ,IAAI,eAAe,QAAQ;AAGxD,KAAI,MAAM,SAAS,KAAA,EAClB,QAAO,KAAK,IAAI,SAAS,aAAa;CAGvC,MAAM,WAAW,KAAK;CAGtB,IAAI,gBAAgB;CACpB,IAAI,aAAa;AACjB,YAAW,MAAM,KAAK,MAAM,OAAO,EAAE;AACpC;AACA,MAAI,EAAE,SAAS,SAAU;;AAI1B,KAAI,kBAAkB,EACrB,QAAO,KAAK,IAAI,SAAS,aAAa;CAIvC,MAAM,QAAQ,eADC,KAAK,IAAI,aAAa,cAAc;AAGnD,QAAO,KAAK,IAAI,SAAS,MAAM;;;;;;;ACxFhC,SAAgB,MACf,OACA,QACA,QACA,QACS;CACT,MAAM,EAAE,UAAU,UAAU,UAAU,EAAE;CAExC,MAAM,EAAE,SAAS,iBAAiB,gBAAA,eAAe,OAAO,QAAQ,OAAO;CAGvE,MAAM,aAAa,MAAM,QAAQ,OAAO;CACxC,MAAM,aAAa,MAAM,QAAQ,OAAO;AAGxC,KAAI,YAAY,SAAS,KAAA,KAAa,YAAY,SAAS,KAAA,EAC1D,QAAO,KAAK,IAAI,SAAS,aAAa;CAIvC,MAAM,kBAAkB,gBAAA,iBAAiB,OAAO,WAAW,KAAK;CAChE,MAAM,kBAAkB,gBAAA,iBAAiB,OAAO,WAAW,KAAK;AAGhE,KAAI,oBAAoB,KAAK,oBAAoB,EAChD,QAAO,KAAK,IAAI,SAAS,aAAa;CAGvC,MAAM,eAAe,KAAK,IAAI,MAAM,YAAY,gBAAgB;CAChE,MAAM,eAAe,KAAK,IAAI,MAAM,YAAY,gBAAgB;CAEhE,MAAM,QAAQ,eAAe,eAAe;AAG5C,QAAO,KAAK,IAAI,SAAS,MAAM;;;;;;;;AAShC,eAAsB,WACrB,OACA,QACA,QACA,QACkB;CAClB,MAAM,EAAE,UAAU,UAAU,UAAU,EAAE;CAGxC,MAAM,CAAC,WAAW,WAAW,YAAY,cAAc,MAAM,QAAQ,IAAI;EACxE,cAAA,qBAAqB,MAAM,WAAW,OAAO,CAAC;EAC9C,cAAA,qBAAqB,MAAM,WAAW,OAAO,CAAC;EAC9C,MAAM,QAAQ,OAAO;EACrB,MAAM,QAAQ,OAAO;EACrB,CAAC;CAEF,MAAM,SAAS,IAAI,IAAI,UAAU,QAAQ,MAAM,MAAM,OAAO,CAAC;CAC7D,MAAM,SAAS,IAAI,IAAI,UAAU,QAAQ,MAAM,MAAM,OAAO,CAAC;CAG7D,IAAI,eAAe;AACnB,MAAK,MAAM,KAAK,OACf,KAAI,OAAO,IAAI,EAAE,CAAE;CAEpB,MAAM,QAAQ,OAAO,OAAO,OAAO,OAAO;CAC1C,MAAM,eAAe,QAAQ,IAAI,eAAe,QAAQ;AAGxD,KAAI,YAAY,SAAS,KAAA,KAAa,YAAY,SAAS,KAAA,EAC1D,QAAO,KAAK,IAAI,SAAS,aAAa;CAGvC,MAAM,aAAa,WAAW;CAC9B,MAAM,aAAa,WAAW;CAG9B,IAAI,aAAa;CACjB,IAAI,kBAAkB;CACtB,IAAI,kBAAkB;AACtB,YAAW,MAAM,UAAU,MAAM,SAAS,EAAE;AAC3C;EACA,MAAM,OAAO,MAAM,MAAM,QAAQ,OAAO;AACxC,MAAI,MAAM,SAAS,WAAY;AAC/B,MAAI,MAAM,SAAS,WAAY;;AAIhC,KAAI,oBAAoB,KAAK,oBAAoB,EAChD,QAAO,KAAK,IAAI,SAAS,aAAa;CAGvC,MAAM,eAAe,KAAK,IAAI,aAAa,gBAAgB;CAC3D,MAAM,eAAe,KAAK,IAAI,aAAa,gBAAgB;CAE3D,MAAM,QAAQ,eAAe,eAAe;AAE5C,QAAO,KAAK,IAAI,SAAS,MAAM;;;;;;;;;;;;;;;;ACvFhC,SAAgB,SACf,OACA,QACA,QACA,QACS;CACT,MAAM,EACL,UAAU,OACV,mBAAmB,IACnB,eAAe,IACf,gBAAgB,OACb,UAAU,EAAE;CAGhB,MAAM,EACL,SAAS,cACT,kBACA,qBACG,gBAAA,eAAe,OAAO,QAAQ,OAAO;CAIzC,MAAM,aACL,iBAAiB,SAAS,KAAK,iBAAiB,SAAS,IACtD,IACA,KAAK,IAAI,SAAS,aAAa;CAGnC,MAAM,kBAAkB,WAAW,OAAO,QAAQ,QAAQ;EACzD;EACA,WAAW;EACX,CAAC;CAGF,IAAI;AACJ,KAAI,iBAAiB,OAAO,KAAK,iBAAiB,OAAO,GAAG;EAC3D,MAAM,EAAE,iBAAiB,gBAAA,iBACxB,kBACA,iBACA;EACD,MAAM,YAAY,KAAK,IAAI,iBAAiB,MAAM,iBAAiB,KAAK;AACxE,YAAU,YAAY,IAAI,eAAe,YAAY;OAErD,WAAU;CAIX,MAAM,cAAc,mBAAmB,eAAe;CAGtD,MAAM,SACJ,mBAAmB,aACnB,eAAe,kBACf,gBAAgB,WACjB;AAED,QAAO,KAAK,IAAI,SAAS,KAAK,IAAI,GAAG,MAAM,CAAC;;;;;;;;AAS7C,eAAsB,cACrB,OACA,QACA,QACA,QACkB;CAClB,MAAM,EACL,UAAU,OACV,mBAAmB,IACnB,eAAe,IACf,gBAAgB,OACb,UAAU,EAAE;CAGhB,MAAM,CAAC,WAAW,WAAW,mBAAmB,MAAM,QAAQ,IAAI;EACjE,cAAA,qBAAqB,MAAM,WAAW,OAAO,CAAC;EAC9C,cAAA,qBAAqB,MAAM,WAAW,OAAO,CAAC;EAC9C,gBAAgB,OAAO,QAAQ,QAAQ;GAAE;GAAS,WAAW;GAAM,CAAC;EACpE,CAAC;CAEF,MAAM,SAAS,IAAI,IAAI,UAAU,QAAQ,MAAM,MAAM,OAAO,CAAC;CAC7D,MAAM,SAAS,IAAI,IAAI,UAAU,QAAQ,MAAM,MAAM,OAAO,CAAC;CAG7D,IAAI,eAAe;AACnB,MAAK,MAAM,KAAK,OACf,KAAI,OAAO,IAAI,EAAE,CAAE;CAEpB,MAAM,QAAQ,OAAO,OAAO,OAAO,OAAO;CAC1C,MAAM,eAAe,QAAQ,IAAI,eAAe,QAAQ;CAGxD,MAAM,aACL,OAAO,SAAS,KAAK,OAAO,SAAS,IAClC,IACA,KAAK,IAAI,SAAS,aAAa;CAGnC,IAAI;AACJ,KAAI,OAAO,OAAO,KAAK,OAAO,OAAO,GAAG;EACvC,MAAM,YAAY,KAAK,IAAI,OAAO,MAAM,OAAO,KAAK;AACpD,YAAU,YAAY,IAAI,eAAe,YAAY;OAErD,WAAU;CAIX,MAAM,cAAc,mBAAmB,eAAe;CAGtD,MAAM,SACJ,mBAAmB,aACnB,eAAe,kBACf,gBAAgB,WACjB;AAED,QAAO,KAAK,IAAI,SAAS,KAAK,IAAI,GAAG,MAAM,CAAC"}