feat: call resolution precision/recall benchmark suite (4.4)

CONTRIBUTING.md

@@ -153,11 +153,12 @@ enabled. The test structure:
 
 ```
 tests/
-  integration/     # buildGraph + full query commands
-  graph/           # Cycle detection, DOT/Mermaid export
-  parsers/         # Language parser extraction (one file per language)
-  search/          # Semantic search + embeddings
-  fixtures/        # Sample projects used by tests
+  integration/          # buildGraph + full query commands
+  graph/                # Cycle detection, DOT/Mermaid export
+  parsers/              # Language parser extraction (one file per language)
+  search/               # Semantic search + embeddings
+  benchmarks/resolution/ # Call resolution precision/recall (per-language fixtures)
+  fixtures/             # Sample projects used by tests
 ```
 
 - Integration tests create temporary copies of fixture projects for isolation
@@ -166,18 +167,45 @@ tests/
 
 ## Regression Benchmarks
 
-Two regression benchmark scripts live in `scripts/`. These are **not** unit
-tests — they measure performance metrics that reviewers use to judge whether a
-change is acceptable. If your PR touches code covered by a benchmark, you
-**must** run it before and after your changes and include the results in the PR
-description.
+Several regression benchmarks track codegraph's accuracy and performance across
+versions. Some live in `scripts/` (run manually), while the resolution benchmark
+runs automatically as part of `npm test`. If your PR touches code covered by a
+benchmark, you **must** run it before and after your changes and include the
+results in the PR description.
 
 | Benchmark | What it measures | When to run |
 |-----------|-----------------|-------------|
 | `node scripts/benchmark.js` | Build speed (native vs WASM), query latency | Changes to `builder.js`, `parser.js`, `queries.js`, `resolve.js`, `db.js`, or the native engine |
 | `node scripts/embedding-benchmark.js` | Search recall (Hit@1/3/5/10) across models | Changes to `embedder.js` or embedding strategies |
 | `node scripts/query-benchmark.js` | Query depth scaling, diff-impact latency | Changes to `queries.js`, `resolve.js`, or `db.js` |
 | `node scripts/incremental-benchmark.js` | Incremental build, import resolution throughput | Changes to `builder.js`, `resolve.js`, `parser.js`, or `journal.js` |
+| `npx vitest run tests/benchmarks/resolution/` | Call resolution precision/recall per language | Changes to `build-edges.js`, `resolve.js`, `parser.js`, or any extractor |
+
+### Resolution precision/recall benchmark
+
+The resolution benchmark (`tests/benchmarks/resolution/`) measures how
+accurately codegraph resolves call edges. It uses hand-annotated fixture projects
+with an `expected-edges.json` manifest per language that declares every call edge
+that should be detected.
+
+The benchmark runner builds the graph for each fixture, compares resolved edges
+against the manifest, and reports:
+
+- **Precision** — what fraction of resolved edges are correct (no false positives)
+- **Recall** — what fraction of expected edges were found (no false negatives)
+- **Per-mode breakdown** — separate recall for `static`, `receiver-typed`, and
+  `interface-dispatched` resolution modes
+
+**CI gate:** The benchmark runs as part of `npm test`. If precision or recall
+drops below the configured thresholds for any language, the test fails.
+
+**Adding a new language fixture:**
+
+1. Create `tests/benchmarks/resolution/fixtures/<language>/` with source files
+2. Add an `expected-edges.json` manifest (see the JSON schema at
+   `tests/benchmarks/resolution/expected-edges.schema.json`)
+3. Add thresholds in `resolution-benchmark.test.js` → `THRESHOLDS`
+4. The benchmark runner auto-discovers fixtures with an `expected-edges.json`
 
 ### How to report results
 

README.md

@@ -546,7 +546,7 @@ Codegraph also extracts symbols from common callback patterns: Commander `.comma
 
 ## 📊 Performance
 
-Self-measured on every release via CI ([build benchmarks](generated/benchmarks/BUILD-BENCHMARKS.md) | [embedding benchmarks](generated/benchmarks/EMBEDDING-BENCHMARKS.md)):
+Self-measured on every release via CI ([build benchmarks](generated/benchmarks/BUILD-BENCHMARKS.md) | [embedding benchmarks](generated/benchmarks/EMBEDDING-BENCHMARKS.md) | [query benchmarks](generated/benchmarks/QUERY-BENCHMARKS.md) | [incremental benchmarks](generated/benchmarks/INCREMENTAL-BENCHMARKS.md) | [resolution precision/recall](tests/benchmarks/resolution/)):
 
 | Metric | Latest |
 |---|---|

tests/benchmarks/resolution/expected-edges.schema.json

@@ -0,0 +1,55 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "title": "Expected Call Edges Manifest",
+  "description": "Hand-annotated call edges for resolution precision/recall benchmarks",
+  "type": "object",
+  "required": ["language", "edges"],
+  "properties": {
+    "language": {
+      "type": "string",
+      "description": "Language identifier matching the fixture directory name"
+    },
+    "description": {
+      "type": "string"
+    },
+    "edges": {
+      "type": "array",
+      "items": {
+        "type": "object",
+        "required": ["source", "target", "kind", "mode"],
+        "properties": {
+          "source": {
+            "type": "object",
+            "required": ["name", "file"],
+            "properties": {
+              "name": { "type": "string", "description": "Function/method name" },
+              "file": { "type": "string", "description": "Filename (basename only)" }
+            }
+          },
+          "target": {
+            "type": "object",
+            "required": ["name", "file"],
+            "properties": {
+              "name": { "type": "string", "description": "Function/method name" },
+              "file": { "type": "string", "description": "Filename (basename only)" }
+            }
+          },
+          "kind": {
+            "type": "string",
+            "enum": ["calls"],
+            "description": "Edge kind — currently only 'calls'"
+          },
+          "mode": {
+            "type": "string",
+            "enum": ["static", "receiver-typed", "interface-dispatched"],
+            "description": "Resolution mode that should produce this edge"
+          },
+          "notes": {
+            "type": "string",
+            "description": "Human-readable explanation of why this edge is expected"
+          }
+        }
+      }
+    }
+  }
+}

tests/benchmarks/resolution/fixtures/javascript/expected-edges.json

@@ -0,0 +1,112 @@
+{
+  "$schema": "../../expected-edges.schema.json",
+  "language": "javascript",
+  "description": "Hand-annotated call edges for JavaScript resolution benchmark",
+  "edges": [
+    {
+      "source": { "name": "UserService.createUser", "file": "service.js" },
+      "target": { "name": "normalize", "file": "validators.js" },
+      "kind": "calls",
+      "mode": "static",
+      "notes": "Direct imported function call from method"
+    },
+    {
+      "source": { "name": "UserService.createUser", "file": "service.js" },
+      "target": { "name": "validate", "file": "validators.js" },
+      "kind": "calls",
+      "mode": "static",
+      "notes": "Direct imported function call from method"
+    },
+    {
+      "source": { "name": "UserService.createUser", "file": "service.js" },
+      "target": { "name": "Logger.error", "file": "logger.js" },
+      "kind": "calls",
+      "mode": "receiver-typed",
+      "notes": "this.logger.error() — receiver-typed via constructor assignment"
+    },
+    {
+      "source": { "name": "UserService.createUser", "file": "service.js" },
+      "target": { "name": "Logger.info", "file": "logger.js" },
+      "kind": "calls",
+      "mode": "receiver-typed",
+      "notes": "this.logger.info() — receiver-typed via constructor assignment"
+    },
+    {
+      "source": { "name": "UserService.deleteUser", "file": "service.js" },
+      "target": { "name": "Logger.warn", "file": "logger.js" },
+      "kind": "calls",
+      "mode": "receiver-typed",
+      "notes": "this.logger.warn() — receiver-typed via constructor assignment"
+    },
+    {
+      "source": { "name": "Logger.info", "file": "logger.js" },
+      "target": { "name": "Logger._write", "file": "logger.js" },
+      "kind": "calls",
+      "mode": "static",
+      "notes": "this._write() — same-class method call"
+    },
+    {
+      "source": { "name": "Logger.warn", "file": "logger.js" },
+      "target": { "name": "Logger._write", "file": "logger.js" },
+      "kind": "calls",
+      "mode": "static",
+      "notes": "this._write() — same-class method call"
+    },
+    {
+      "source": { "name": "Logger.error", "file": "logger.js" },
+      "target": { "name": "Logger._write", "file": "logger.js" },
+      "kind": "calls",
+      "mode": "static",
+      "notes": "this._write() — same-class method call"
+    },
+    {
+      "source": { "name": "validate", "file": "validators.js" },
+      "target": { "name": "checkLength", "file": "validators.js" },
+      "kind": "calls",
+      "mode": "static",
+      "notes": "Same-file function call"
+    },
+    {
+      "source": { "name": "normalize", "file": "validators.js" },
+      "target": { "name": "trimWhitespace", "file": "validators.js" },
+      "kind": "calls",
+      "mode": "static",
+      "notes": "Same-file function call"
+    },
+    {
+      "source": { "name": "main", "file": "index.js" },
+      "target": { "name": "buildService", "file": "service.js" },
+      "kind": "calls",
+      "mode": "static",
+      "notes": "Direct imported function call"
+    },
+    {
+      "source": { "name": "main", "file": "index.js" },
+      "target": { "name": "UserService.createUser", "file": "service.js" },
+      "kind": "calls",
+      "mode": "receiver-typed",
+      "notes": "svc.createUser() — receiver typed via buildService() return"
+    },
+    {
+      "source": { "name": "main", "file": "index.js" },
+      "target": { "name": "validate", "file": "validators.js" },
+      "kind": "calls",
+      "mode": "static",
+      "notes": "Direct imported function call"
+    },
+    {
+      "source": { "name": "main", "file": "index.js" },
+      "target": { "name": "UserService.deleteUser", "file": "service.js" },
+      "kind": "calls",
+      "mode": "receiver-typed",
+      "notes": "svc.deleteUser() — receiver typed via buildService() return"
+    },
+    {
+      "source": { "name": "directInstantiation", "file": "index.js" },
+      "target": { "name": "UserService.createUser", "file": "service.js" },
+      "kind": "calls",
+      "mode": "receiver-typed",
+      "notes": "svc.createUser() — receiver typed via new UserService()"
+    }
+  ]
+}

tests/benchmarks/resolution/fixtures/javascript/index.js

@@ -0,0 +1,15 @@
+import { buildService, UserService } from './service.js';
+import { validate } from './validators.js';
+
+export function main() {
+  const svc = buildService();
+  const result = svc.createUser({ name: 'Alice' });
+  if (result && validate(result)) {
+    svc.deleteUser(1);
+  }
+}
+
+export function directInstantiation() {
+  const svc = new UserService();
+  return svc.createUser({ name: 'Bob' });
+}

tests/benchmarks/resolution/fixtures/javascript/logger.js

@@ -0,0 +1,21 @@
+export class Logger {
+  constructor(prefix) {
+    this.prefix = prefix;
+  }
+
+  info(msg) {
+    this._write('INFO', msg);
+  }
+
+  warn(msg) {
+    this._write('WARN', msg);
+  }
+
+  error(msg) {
+    this._write('ERROR', msg);
+  }
+
+  _write(level, msg) {
+    console.log(`[${this.prefix}] ${level}: ${msg}`);
+  }
+}

tests/benchmarks/resolution/fixtures/javascript/service.js

@@ -0,0 +1,27 @@
+import { Logger } from './logger.js';
+import { normalize, validate } from './validators.js';
+
+export class UserService {
+  constructor() {
+    this.logger = new Logger('UserService');
+  }
+
+  createUser(data) {
+    const clean = normalize(data);
+    if (!validate(clean)) {
+      this.logger.error('Validation failed');
+      return null;
+    }
+    this.logger.info('User created');
+    return clean;
+  }
+
+  deleteUser(id) {
+    this.logger.warn(`Deleting user ${id}`);
+    return true;
+  }
+}
+
+export function buildService() {
+  return new UserService();
+}

tests/benchmarks/resolution/fixtures/javascript/validators.js

@@ -0,0 +1,15 @@
+export function validate(data) {
+  return data != null && typeof data.name === 'string' && checkLength(data.name);
+}
+
+export function normalize(data) {
+  return { ...data, name: trimWhitespace(data.name) };
+}
+
+function checkLength(str) {
+  return str.length > 0 && str.length < 256;
+}
+
+function trimWhitespace(str) {
+  return str.trim();
+}