From 7bd9eb5308eb6201515ff03f8007c512812824aa Mon Sep 17 00:00:00 2001
From: Giancarlo Erra <giancarlo@widescreen.studio>
Date: Thu, 11 Jun 2026 12:12:18 +0100
Subject: [PATCH 1/2] feat(graph): full Dart support via tree-sitter AST (#71)

Dart previously fell through to the regex symbol fallback, which cannot
match type-first signatures (void foo(), Future<int> baz() async), so
classes, methods, and calls were invisible to codebase_symbol,
codebase_flow, and codebase_impact, and files were chunked by line
count instead of declaration boundaries.

- Register @ast-grep/lang-dart as a dynamic grammar (per-grammar
  failure isolation keeps missing prebuilds on the regex fallback).
- Add extractFromDart: classes, mixins (trait), enums, extensions,
  typedefs, type-first top-level functions, getters/setters, and
  constructors including named and factory forms. Scope ranges are
  stitched from Dart's sibling function_signature/function_body pairs.
  Calls are recovered from argument_part nodes: method calls, bare
  calls, constructor invocations, prefixed calls, and cascades.
- Add dart to TOP_LEVEL_KINDS so chunking follows declaration
  boundaries; signature and body kinds are both listed so the
  overlap-merge fuses each pair into one region.
- Add dart main() to ENTRY_POINT_NAMES for entry-point detection.
- Move Dart to Full Support in the README language matrix.
---
 README.md                            |   6 +-
 package-lock.json                    |  19 +++
 package.json                         |   1 +
 src/constants.ts                     |   1 +
 src/services/code-graph.ts           |   1 +
 src/services/graph-symbols.ts        | 210 ++++++++++++++++++++++++++-
 src/services/indexer.ts              |   5 +
 tests/unit/graph-entrypoints.test.ts |  22 +++
 tests/unit/graph-symbols.test.ts     | 120 +++++++++++++--
 tests/unit/indexer.test.ts           |  42 ++++++
 10 files changed, 413 insertions(+), 14 deletions(-)

diff --git a/README.md b/README.md
index b5035fe..b9948e9 100644
--- a/README.md
+++ b/README.md
@@ -959,13 +959,15 @@ SocratiCode supports languages at three levels:
 
 ### Full Support (indexing + code graph + AST chunking)
 
-JavaScript, TypeScript, TSX, Python, Java, Kotlin, Scala, C, C++, C#, Go, Rust, Ruby, PHP, Swift, Bash/Shell, HTML, CSS/SCSS, Svelte, Vue
+JavaScript, TypeScript, TSX, Python, Java, Kotlin, Scala, C, C++, C#, Go, Rust, Ruby, PHP, Swift, Dart, Bash/Shell, HTML, CSS/SCSS, Svelte, Vue
 
 Svelte and Vue: imports extracted from `<script>` blocks (re-parsed as TypeScript) and CSS `@import`/`@require` from `<style>` blocks (any combination of `lang`, `scoped`, `module`, `global` attributes). Path aliases from `tsconfig.json`/`jsconfig.json` `compilerOptions.paths` are resolved (including `extends` chains). SCSS partial resolution (`_` prefix convention) is supported.
 
+Dart: symbols (classes, mixins, enums, extensions, typedefs, functions, getters/setters, constructors including named and factory), call sites (method calls, cascades, constructor invocations), `main()` entry-point detection, and AST chunking are all tree-sitter based; import/export/part edges are extracted via regex.
+
 ### Code Graph via Regex + Indexing
 
-Dart (import/export/part), Lua (require/dofile/loadfile), SASS, LESS (CSS `@import` extraction)
+Lua (require/dofile/loadfile), SASS, LESS (CSS `@import` extraction)
 
 ### Indexing Only (hybrid search, line-based chunking)
 
diff --git a/package-lock.json b/package-lock.json
index e3491c2..04941ce 100644
--- a/package-lock.json
+++ b/package-lock.json
@@ -13,6 +13,7 @@
         "@ast-grep/lang-c": "^0.0.5",
         "@ast-grep/lang-cpp": "^0.0.5",
         "@ast-grep/lang-csharp": "^0.0.5",
+        "@ast-grep/lang-dart": "^0.0.7",
         "@ast-grep/lang-go": "^0.0.5",
         "@ast-grep/lang-java": "^0.0.6",
         "@ast-grep/lang-kotlin": "^0.0.6",
@@ -126,6 +127,24 @@
         }
       }
     },
+    "node_modules/@ast-grep/lang-dart": {
+      "version": "0.0.7",
+      "resolved": "https://registry.npmjs.org/@ast-grep/lang-dart/-/lang-dart-0.0.7.tgz",
+      "integrity": "sha512-9tTkOnw10g+/cgt8jw1XqZ0AKLDXoKYP1HoY3reZPQLhk4IWs8mNAu4Yd6E80CdQ2sJ10+L7wv1C16jTt4Yirw==",
+      "hasInstallScript": true,
+      "license": "ISC",
+      "dependencies": {
+        "@ast-grep/setup-lang": "0.0.6"
+      },
+      "peerDependencies": {
+        "tree-sitter-cli": "0.25.8"
+      },
+      "peerDependenciesMeta": {
+        "tree-sitter-cli": {
+          "optional": true
+        }
+      }
+    },
     "node_modules/@ast-grep/lang-go": {
       "version": "0.0.5",
       "resolved": "https://registry.npmjs.org/@ast-grep/lang-go/-/lang-go-0.0.5.tgz",
diff --git a/package.json b/package.json
index efafb17..71e45ab 100644
--- a/package.json
+++ b/package.json
@@ -67,6 +67,7 @@
     "@ast-grep/lang-c": "^0.0.5",
     "@ast-grep/lang-cpp": "^0.0.5",
     "@ast-grep/lang-csharp": "^0.0.5",
+    "@ast-grep/lang-dart": "^0.0.7",
     "@ast-grep/lang-go": "^0.0.5",
     "@ast-grep/lang-java": "^0.0.6",
     "@ast-grep/lang-kotlin": "^0.0.6",
diff --git a/src/constants.ts b/src/constants.ts
index 5f84659..07c1713 100644
--- a/src/constants.ts
+++ b/src/constants.ts
@@ -163,6 +163,7 @@ export const ENTRY_POINT_NAMES: Record<string, Set<string>> = {
   swift: new Set(["main"]),
   ruby: new Set(["main"]),
   php: new Set(["main"]),
+  dart: new Set(["main"]),
 };
 
 // ── Path normalization ──────────────────────────────────────────────────
diff --git a/src/services/code-graph.ts b/src/services/code-graph.ts
index e656c2c..8d7cbdd 100644
--- a/src/services/code-graph.ts
+++ b/src/services/code-graph.ts
@@ -493,6 +493,7 @@ export function ensureDynamicLanguages(): void {
       ["bash",    "@ast-grep/lang-bash"],
       ["php",     "@ast-grep/lang-php"],
       ["lua",     "@ast-grep/lang-lua"],
+      ["dart",    "@ast-grep/lang-dart"],
     ];
 
     for (const [name, pkg] of langPackages) {
diff --git a/src/services/graph-symbols.ts b/src/services/graph-symbols.ts
index e0b9878..4c7f022 100644
--- a/src/services/graph-symbols.ts
+++ b/src/services/graph-symbols.ts
@@ -145,7 +145,10 @@ export function extractSymbolsAndCalls(
     if (langKey === "lua") {
       return extractFromLua(source, relativePath, language, moduleSymbol);
     }
-    // Dart, Svelte, Vue and others fall through to the regex fallback.
+    if (langKey === "dart") {
+      return extractFromDart(source, relativePath, language, moduleSymbol);
+    }
+    // Svelte, Vue and others fall through to the regex fallback.
     return extractFromRegex(source, relativePath, language, moduleSymbol);
   } catch (err) {
     if (!symbolExtractionWarned.has(langKey)) {
@@ -276,6 +279,211 @@ function extractFromLua(
   return { symbols, rawCalls };
 }
 
+// ── Dart (type-first signatures, sibling signature/body pairs, selector calls) ──
+
+/**
+ * Dart previously fell through to the regex fallback, which cannot match
+ * type-first signatures (`void foo()`, `Future<int> baz() async`), so
+ * classes, methods, and call sites were invisible to the symbol graph.
+ * This walks the ast-grep Dart tree instead. Grammar quirks handled here:
+ * class/mixin/enum/extension nodes span their bodies, but a function is a
+ * `function_signature` followed by a SIBLING `function_body`, so scope
+ * ranges are stitched from each pair; plain constructors live inside a
+ * generic `declaration` wrapper; and there is no call_expression kind, so
+ * calls are recovered from `argument_part` nodes (callee = the preceding
+ * identifier or selector chain, or the `cascade_selector` for `..` calls).
+ */
+function extractFromDart(
+  source: string,
+  file: string,
+  language: string,
+  moduleSym: SymbolNode,
+): ExtractedSymbols {
+  const root = parse("dart" as unknown as Lang, source).root();
+  const symbols: SymbolNode[] = [moduleSym];
+  const scopes: ScopeFrame[] = [];
+
+  // biome-ignore lint/suspicious/noExplicitAny: ast-grep node type leaks through
+  const kidsOf = (n: any): any[] => {
+    try {
+      return n.children();
+    } catch {
+      return [];
+    }
+  };
+  // biome-ignore lint/suspicious/noExplicitAny: ast-grep node type leaks through
+  const childOfKind = (n: any, kind: string): any | null =>
+    // biome-ignore lint/suspicious/noExplicitAny: ast-grep node type leaks through
+    kidsOf(n).find((c: any) => c.kind() === kind) ?? null;
+  // Direct identifier children only — the name slot. Type annotations are
+  // `type_identifier`/`void_type` and parameter names are nested deeper, so
+  // they never appear here.
+  // biome-ignore lint/suspicious/noExplicitAny: ast-grep node type leaks through
+  const idChildren = (n: any): any[] =>
+    // biome-ignore lint/suspicious/noExplicitAny: ast-grep node type leaks through
+    kidsOf(n).filter((c: any) => c.kind() === "identifier");
+
+  const addSym = (
+    name: string,
+    qualifiedName: string,
+    kind: SymbolKind,
+    startLine: number,
+    endLine: number,
+  ): void => {
+    const sym: SymbolNode = {
+      id: makeId(file, qualifiedName, startLine),
+      name,
+      qualifiedName,
+      kind,
+      file,
+      line: startLine,
+      endLine,
+      language,
+    };
+    symbols.push(sym);
+    scopes.push({ name: qualifiedName, startLine, endLine, symbolId: sym.id });
+  };
+
+  // biome-ignore lint/suspicious/noExplicitAny: ast-grep node type leaks through
+  const lineOf = (n: any): number => n.range().start.line + 1;
+  // biome-ignore lint/suspicious/noExplicitAny: ast-grep node type leaks through
+  const endLineOf = (n: any): number => n.range().end.line + 1;
+
+  /**
+   * Emit the member symbols of a class-like body. Members come in ordered
+   * sibling pairs: a `method_signature` (wrapping function/getter/setter/
+   * factory signatures) or a `declaration` (fields and plain constructors),
+   * optionally followed by its `function_body`.
+   */
+  // biome-ignore lint/suspicious/noExplicitAny: ast-grep node type leaks through
+  const walkMembers = (bodyNode: any, owner: string): void => {
+    const members = kidsOf(bodyNode);
+    for (let i = 0; i < members.length; i++) {
+      const member = members[i];
+      const memberKind = member.kind();
+      const next = members[i + 1];
+      const scopeEnd = next && next.kind() === "function_body" ? endLineOf(next) : endLineOf(member);
+
+      if (memberKind === "method_signature") {
+        const inner = kidsOf(member)[0];
+        if (!inner) continue;
+        const innerKind = inner.kind();
+        if (innerKind === "factory_constructor_signature") {
+          const ids = idChildren(inner);
+          if (ids.length === 0) continue;
+          // biome-ignore lint/suspicious/noExplicitAny: ast-grep node type leaks through
+          const qn = ids.map((c: any) => c.text()).join(".");
+          addSym(ids[ids.length - 1].text(), qn, "constructor", lineOf(member), scopeEnd);
+        } else if (
+          innerKind === "function_signature" ||
+          innerKind === "getter_signature" ||
+          innerKind === "setter_signature"
+        ) {
+          const ids = idChildren(inner);
+          if (ids.length === 0) continue;
+          const name = ids[ids.length - 1].text();
+          addSym(name, `${owner}.${name}`, "method", lineOf(member), scopeEnd);
+        }
+      } else if (memberKind === "declaration") {
+        // Plain (possibly named) constructors: `Foo(this.c);` / `Foo.named(...)`.
+        // Field declarations have no constructor_signature child and are skipped.
+        const ctor = childOfKind(member, "constructor_signature");
+        if (!ctor) continue;
+        const ids = idChildren(ctor);
+        if (ids.length === 0) continue;
+        // biome-ignore lint/suspicious/noExplicitAny: ast-grep node type leaks through
+        const qn = ids.map((c: any) => c.text()).join(".");
+        addSym(ids[ids.length - 1].text(), qn, "constructor", lineOf(member), scopeEnd);
+      }
+    }
+  };
+
+  // ── Top-level declarations (ordered walk so signature/body pairs line up) ──
+  const topLevel = kidsOf(root);
+  for (let i = 0; i < topLevel.length; i++) {
+    const node = topLevel[i];
+    const nodeKind = node.kind();
+
+    if (nodeKind === "class_definition" || nodeKind === "mixin_declaration" || nodeKind === "extension_declaration") {
+      const nameNode = childOfKind(node, "identifier");
+      if (!nameNode) continue;
+      const name = nameNode.text();
+      const kind: SymbolKind = nodeKind === "mixin_declaration" ? "trait" : "class";
+      addSym(name, name, kind, lineOf(node), endLineOf(node));
+      const body = childOfKind(node, "class_body") ?? childOfKind(node, "extension_body");
+      if (body) walkMembers(body, name);
+    } else if (nodeKind === "enum_declaration") {
+      const nameNode = childOfKind(node, "identifier");
+      if (nameNode) addSym(nameNode.text(), nameNode.text(), "enum", lineOf(node), endLineOf(node));
+    } else if (nodeKind === "type_alias") {
+      const nameNode = childOfKind(node, "type_identifier");
+      if (nameNode) addSym(nameNode.text(), nameNode.text(), "interface", lineOf(node), endLineOf(node));
+    } else if (nodeKind === "function_signature" || nodeKind === "getter_signature" || nodeKind === "setter_signature") {
+      const ids = idChildren(node);
+      if (ids.length === 0) continue;
+      const name = ids[ids.length - 1].text();
+      const next = topLevel[i + 1];
+      const scopeEnd = next && next.kind() === "function_body" ? endLineOf(next) : endLineOf(node);
+      addSym(name, name, "function", lineOf(node), scopeEnd);
+    }
+  }
+
+  // ── Calls — every invocation wraps an `argument_part` node ──────────────
+  const rawCalls: ExtractedSymbols["rawCalls"] = [];
+  for (const ap of safeFindAll(root, "argument_part")) {
+    const holder = ap.parent();
+    if (!holder) continue;
+    const holderKind = holder.kind();
+    let callee: string | null = null;
+
+    if (holderKind === "cascade_section") {
+      // `obj..method(args)` — the callee lives in the cascade_selector.
+      const cs = childOfKind(holder, "cascade_selector");
+      const id = cs ? childOfKind(cs, "identifier") : null;
+      callee = id ? id.text() : null;
+    } else if (holderKind === "selector") {
+      // `name(args)` / `expr.name(args)` — the callee is the previous
+      // sibling: a bare identifier, or a selector whose trailing identifier
+      // is the method name (`f.bar(…)`, `mat.runApp(…)`, `Foo.create(…)`).
+      const parent = holder.parent();
+      if (!parent) continue;
+      const siblings = kidsOf(parent);
+      const hr = holder.range();
+      const idx = siblings.findIndex(
+        // biome-ignore lint/suspicious/noExplicitAny: ast-grep node type leaks through
+        (c: any) => {
+          if (c.kind() !== "selector") return false;
+          const r = c.range();
+          return (
+            r.start.line === hr.start.line &&
+            r.start.column === hr.start.column &&
+            r.end.line === hr.end.line &&
+            r.end.column === hr.end.column
+          );
+        },
+      );
+      if (idx <= 0) continue;
+      const prev = siblings[idx - 1];
+      if (prev.kind() === "identifier") {
+        callee = prev.text();
+      } else if (prev.kind() === "selector") {
+        const ids = safeFindAll(prev, "identifier");
+        callee = ids.length > 0 ? ids[ids.length - 1].text() : null;
+      }
+    }
+
+    if (!callee) continue;
+    const line = ap.range().start.line + 1;
+    rawCalls.push({
+      callerId: findCallerId(scopes, line, moduleSym.id),
+      calleeName: callee,
+      callSite: { file, line },
+    });
+  }
+
+  return { symbols, rawCalls };
+}
+
 // ── JS / TS / TSX ────────────────────────────────────────────────────────
 
 function extractFromTsLike(
diff --git a/src/services/indexer.ts b/src/services/indexer.ts
index 9579d4d..d5137cb 100644
--- a/src/services/indexer.ts
+++ b/src/services/indexer.ts
@@ -272,6 +272,11 @@ const TOP_LEVEL_KINDS: Record<string, string[]> = {
   php:        ["function_definition", "class_declaration", "method_declaration", "trait_declaration"],
   swift:      ["function_declaration", "class_declaration", "struct_declaration", "protocol_declaration", "extension_declaration"],
   bash:       ["function_definition"],
+  // Dart: class/mixin/enum/extension nodes span their bodies, but a top-level
+  // function is a `function_signature` followed by a SIBLING `function_body`
+  // starting on the same line. Both kinds are listed so the overlap-merge in
+  // findAstBoundaries fuses each signature/body pair into one region.
+  dart:       ["class_definition", "mixin_declaration", "enum_declaration", "extension_declaration", "type_alias", "function_signature", "function_body"],
 };
 
 /** Minimum lines for a chunk to stand on its own (otherwise merge with neighbors) */
diff --git a/tests/unit/graph-entrypoints.test.ts b/tests/unit/graph-entrypoints.test.ts
index 3276ece..393fddc 100644
--- a/tests/unit/graph-entrypoints.test.ts
+++ b/tests/unit/graph-entrypoints.test.ts
@@ -96,6 +96,28 @@ describe("graph-entrypoints", () => {
     expect(entries.some((e) => e.name === "main")).toBe(true);
   });
 
+  it("detects Dart main() as a conventional entry point", () => {
+    const graph: CodeGraph = {
+      nodes: [
+        {
+          relativePath: "lib/main.dart",
+          imports: [],
+          exports: [],
+          dependencies: [],
+          dependents: ["lib/app.dart"], // not an orphan — name heuristic must fire
+        },
+      ],
+      edges: [],
+    };
+    const payloads: SymbolGraphFilePayload[] = [
+      { ...mkPayload("lib/main.dart", [{ name: "main", line: 3 }]), language: "dart" },
+    ];
+    const entries = detectEntryPoints(graph, payloads);
+    const main = entries.find((e) => e.name === "main");
+    expect(main).toBeDefined();
+    expect(main?.reason).toBe("well-known-name:main");
+  });
+
   it("returns empty array when nothing matches", () => {
     const graph: CodeGraph = { nodes: [], edges: [] };
     const entries = detectEntryPoints(graph, []);
diff --git a/tests/unit/graph-symbols.test.ts b/tests/unit/graph-symbols.test.ts
index cd6c47c..aa75c7b 100644
--- a/tests/unit/graph-symbols.test.ts
+++ b/tests/unit/graph-symbols.test.ts
@@ -472,25 +472,123 @@ end
     });
   });
 
-  describe("Regex fallback (Dart, Svelte, Vue, unknown)", () => {
-    it("handles Dart via regex fallback", () => {
+  describe("Dart", () => {
+    it("extracts type-first declarations the regex fallback could never match", () => {
       const src = `
-String greet(String name) {
-  return 'Hi $name';
+class Foo {
+  Foo(int c);
+  Foo.named(int c);
+  factory Foo.create() => Foo(1);
+  void bar(int x) {
+    helper(x);
+  }
+  String get name => 'foo';
+  set name(String v) {}
+}
+
+mixin Loggable {
+  void log(String msg) {}
+}
+
+enum Color { red, green }
+
+extension StrExt on String {
+  int len() => length;
+}
+
+typedef Callback = void Function(int);
+
+Future<int> fetchCount() async {
+  return 1;
+}
+`;
+      const out = extractSymbolsAndCalls(src, "dart" as unknown as Lang, ".dart", "lib/foo.dart");
+      const has = (qn: string, kind: string) =>
+        out.symbols.some((s) => s.qualifiedName === qn && s.kind === kind);
+
+      expect(has("Foo", "class")).toBe(true);
+      // Constructors: plain, named, and factory all resolve as constructors
+      // with dotted qualified names — the regex fallback saw none of these.
+      // The plain constructor deliberately shares the class's qualified name
+      // (that is how call sites reference it); they differ by kind and line.
+      expect(has("Foo", "constructor")).toBe(true);
+      expect(has("Foo.named", "constructor")).toBe(true);
+      expect(has("Foo.create", "constructor")).toBe(true);
+      expect(has("Foo.bar", "method")).toBe(true);
+      // Getter and setter share a name but live on different lines (distinct ids)
+      const nameSyms = out.symbols.filter((s) => s.qualifiedName === "Foo.name");
+      expect(nameSyms).toHaveLength(2);
+      expect(has("Loggable", "trait")).toBe(true);
+      expect(has("Loggable.log", "method")).toBe(true);
+      expect(has("Color", "enum")).toBe(true);
+      expect(has("StrExt.len", "method")).toBe(true);
+      expect(has("Callback", "interface")).toBe(true);
+      // Type-first top-level signature (`Future<int> fetchCount()`)
+      expect(has("fetchCount", "function")).toBe(true);
+    });
+
+    it("stitches a top-level function's scope from its sibling signature and body", () => {
+      const src = `
+void helper(int x) {
+  print(x);
+  print(x + 1);
 }
+`;
+      const out = extractSymbolsAndCalls(src, "dart" as unknown as Lang, ".dart", "lib/h.dart");
+      const helper = out.symbols.find((s) => s.qualifiedName === "helper");
+      // The signature node alone ends on line 2; the scope must reach the
+      // body's closing brace, otherwise calls inside attribute to <module>.
+      expect(helper?.line).toBe(2);
+      expect(helper?.endLine).toBe(5);
+      const printCall = out.rawCalls.find((c) => c.calleeName === "print");
+      expect(printCall?.callerId).toContain("::helper#");
+    });
 
+    it("attributes method calls, cascades, and constructor invocations to the enclosing scope", () => {
+      const src = `
 class Foo {
-  int bar() => 1;
+  void bar(int x) {}
+  Future<void> load() async {}
+}
+
+void main() {
+  final f = Foo(1);
+  f.bar(2);
+  f..bar(3)..load();
+  mat.runApp();
+  helper(5);
 }
 `;
-      const out = extractSymbolsAndCalls(src, "dart" as unknown as Lang, ".dart", "main.dart");
-      // regex fallback should at least produce <module> and not throw
-      expect(out.symbols.some((s) => s.name === "<module>")).toBe(true);
-      const names = out.symbols.map((s) => s.name);
-      // best-effort detection: should find at least one named symbol
-      expect(names.length).toBeGreaterThanOrEqual(1);
+      const out = extractSymbolsAndCalls(src, "dart" as unknown as Lang, ".dart", "lib/main.dart");
+      const fromMain = out.rawCalls.filter((c) => c.callerId.includes("::main#"));
+      const callees = fromMain.map((c) => c.calleeName);
+
+      // Constructor invocation (`Foo(1)`, no `new` keyword in modern Dart)
+      expect(callees).toContain("Foo");
+      // Plain method call `f.bar(2)`
+      expect(callees).toContain("bar");
+      // Cascade `..load()` — a Dart-only form with its own grammar shape
+      expect(callees).toContain("load");
+      // Prefixed call `mat.runApp()` resolves to the trailing identifier
+      expect(callees).toContain("runApp");
+      // Bare call
+      expect(callees).toContain("helper");
     });
 
+    it("detects main() so Dart apps get a conventional entry point", () => {
+      const src = `
+void main() {
+  runApp();
+}
+`;
+      const out = extractSymbolsAndCalls(src, "dart" as unknown as Lang, ".dart", "bin/app.dart");
+      const main = out.symbols.find((s) => s.name === "main");
+      expect(main).toBeDefined();
+      expect(main?.kind).toBe("function");
+    });
+  });
+
+  describe("Regex fallback (Svelte, Vue, unknown)", () => {
     it("handles unknown language without throwing", () => {
       const src = "some random text\nwith no recognizable structure";
       const out = extractSymbolsAndCalls(
diff --git a/tests/unit/indexer.test.ts b/tests/unit/indexer.test.ts
index 49b5c57..005b4eb 100644
--- a/tests/unit/indexer.test.ts
+++ b/tests/unit/indexer.test.ts
@@ -188,6 +188,48 @@ describe("indexer utilities", () => {
       }
     });
 
+    it("uses AST-aware chunking for Dart files (class and signature/body boundaries)", () => {
+      // Two large Dart classes plus a top-level function, exceeding CHUNK_SIZE
+      // lines. Dart splits a top-level function into sibling signature and
+      // body nodes; if the two were not merged into one region, the function
+      // body would be severed from its signature.
+      const method = (name: string) => {
+        const body = Array.from({ length: 12 }, (_, j) => `    final v${j} = ${j};`).join("\n");
+        return `  void ${name}() {\n${body}\n  }`;
+      };
+      // Six 14-line methods per class (~86 lines each): two classes exceed the
+      // 150-line merge cap, so the chunker must flush them as separate chunks
+      // instead of grouping them (small adjacent declarations merge by design).
+      const classOf = (name: string) =>
+        `class ${name} {\n${Array.from({ length: 6 }, (_, i) => method(`m${i}`)).join("\n")}\n}`;
+      const topFn = `Future<int> compute() async {\n${Array.from({ length: 12 }, (_, j) => `  final t${j} = ${j};`).join("\n")}\n  return 1;\n}`;
+      const content = `import 'package:foo/bar.dart';\n\n${classOf("Alpha")}\n\n${classOf("Beta")}\n\n${topFn}\n`;
+
+      const chunks = chunkFileContent("/test/app.dart", "app.dart", content);
+
+      expect(chunks.length).toBeGreaterThan(1);
+      for (const chunk of chunks) {
+        expect(chunk.language).toBe("dart");
+      }
+      // AST-aware (not line-based): each class lives in a single chunk whose
+      // content begins at the declaration. Beta spans lines ~90-175, so a
+      // fixed 100-line window would split it mid-body.
+      const lines = content.split("\n");
+      const alphaLine = lines.findIndex((l) => l.startsWith("class Alpha")) + 1;
+      expect(chunks.some((c) => c.startLine === alphaLine)).toBe(true);
+      const betaLine = lines.findIndex((l) => l.startsWith("class Beta")) + 1;
+      const betaClassLines = classOf("Beta").split("\n").length;
+      const betaChunk = chunks.find((c) => c.startLine <= betaLine && c.endLine >= betaLine);
+      expect(betaChunk).toBeDefined();
+      expect(betaChunk?.content.trim().startsWith("class Beta")).toBe(true);
+      expect(betaChunk?.endLine).toBeGreaterThanOrEqual(betaLine + betaClassLines - 1);
+      // The top-level function's signature and body stay in the same chunk.
+      const computeLine = lines.findIndex((l) => l.startsWith("Future<int> compute")) + 1;
+      const computeChunk = chunks.find((c) => c.startLine <= computeLine && c.endLine >= computeLine);
+      expect(computeChunk).toBeDefined();
+      expect(computeChunk?.content).toContain("return 1;");
+    });
+
     it("sets correct language for Python files", () => {
       const content = "def hello():\n    print('hello')";
       const chunks = chunkFileContent("/test/app.py", "app.py", content);

From c5a706cc718a3b0743a5451e7c735df9c6cfbfea Mon Sep 17 00:00:00 2001
From: Giancarlo Erra <giancarlo@widescreen.studio>
Date: Thu, 11 Jun 2026 12:22:15 +0100
Subject: [PATCH 2/2] test(graph): pin graceful degradation for Dart extension
 types

The vendored @ast-grep/lang-dart 0.0.7 grammar predates Dart 3.3
extension types and parses them to ERROR nodes; the kinds a newer
tree-sitter-dart exposes (extension_type_declaration) do not exist in
this version, so matching on them would be unreachable code. Document
the limitation at the walk site and lock in the contract: no throw, no
bogus symbols from the ERROR region, and the rest of the file still
extracts normally.
---
 src/services/graph-symbols.ts    |  5 +++++
 tests/unit/graph-symbols.test.ts | 20 ++++++++++++++++++++
 2 files changed, 25 insertions(+)

diff --git a/src/services/graph-symbols.ts b/src/services/graph-symbols.ts
index 4c7f022..b4b6f5c 100644
--- a/src/services/graph-symbols.ts
+++ b/src/services/graph-symbols.ts
@@ -399,6 +399,11 @@ function extractFromDart(
   };
 
   // ── Top-level declarations (ordered walk so signature/body pairs line up) ──
+  // Dart 3.3 `extension type` is NOT handled: the vendored grammar
+  // (@ast-grep/lang-dart 0.0.7) predates the syntax and parses it to ERROR
+  // nodes (no extension_type_declaration kind exists), so such declarations
+  // degrade to "not extracted" while the rest of the file extracts normally.
+  // Revisit when the upstream grammar adds the kind.
   const topLevel = kidsOf(root);
   for (let i = 0; i < topLevel.length; i++) {
     const node = topLevel[i];
diff --git a/tests/unit/graph-symbols.test.ts b/tests/unit/graph-symbols.test.ts
index aa75c7b..0ab8f83 100644
--- a/tests/unit/graph-symbols.test.ts
+++ b/tests/unit/graph-symbols.test.ts
@@ -575,6 +575,26 @@ void main() {
       expect(callees).toContain("helper");
     });
 
+    it("degrades gracefully on Dart 3.3 extension types (unsupported by grammar 0.0.7)", () => {
+      // The vendored tree-sitter grammar predates `extension type` and parses
+      // it to ERROR nodes. The contract: no throw, no bogus symbols from the
+      // ERROR region, and the rest of the file still extracts normally.
+      const src = `
+extension type Meters(int value) {
+  int get inKm => value ~/ 1000;
+}
+
+class Real {
+  void work() {}
+}
+`;
+      const out = extractSymbolsAndCalls(src, "dart" as unknown as Lang, ".dart", "lib/m.dart");
+      expect(out.symbols.some((s) => s.qualifiedName === "Real" && s.kind === "class")).toBe(true);
+      expect(out.symbols.some((s) => s.qualifiedName === "Real.work" && s.kind === "method")).toBe(true);
+      // The unsupported declaration produces no symbol named Meters
+      expect(out.symbols.some((s) => s.name === "Meters")).toBe(false);
+    });
+
     it("detects main() so Dart apps get a conventional entry point", () => {
       const src = `
 void main() {