fix(index): preserve symbol word matches through full reindexing

CHANGELOG.md

@@ -14,6 +14,8 @@ and adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
 ### Fixes
 
+- Repeating a full index through the library keeps unchanged symbols discoverable by the words in their names, and cancelling a re-index preserves those matches.
+
 - In TypeScript, an awaited call with type arguments like `await fetchApi<User>("/me")` or `await api.get<T>(url)` now links to the function it calls. Before, the call was dropped, so typed fetch wrappers looked like they had no callers. (#2586)
 - In React, a component rendered at module scope, in a top-level constant or inside a test's `it(…)` callback, like `createRoot(el).render(<App />)` or `render(<Card />)`, now counts as rendered there, so an app's entry reaches its root component and a component's tests show among its callers. (#2587)
 - In Python, `import a.b as X` written inside a function now links its calls like `X.f()`, as the same import at module level already did. Thanks @Egrn for the report. (#2600)

__tests__/segment-vocab.test.ts

@@ -73,6 +73,43 @@ export function writeConfig(): void {}
     expect(hit.kind).not.toBe('file');
   });
 
+  it('keeps unchanged symbol words through repeated full indexing', async () => {
+    const before = cg.getSegmentMatches(['state', 'machine']);
+    expect(before.map((m) => m.name)).toContain('OrderStateMachine');
+
+    for (let run = 0; run < 2; run++) {
+      expect((await cg.indexAll()).success).toBe(true);
+      expect(cg.getNodesByName('OrderStateMachine')).toHaveLength(1);
+      expect(cg.getSegmentMatches(['state', 'machine'])).toEqual(before);
+      expect(cg.getSegmentMatches(['checkout', 'service']).map((m) => m.name)).toContain('CheckoutService');
+    }
+  });
+
+  it('rebuilds words for both changed and unchanged files after a full index', async () => {
+    const touched = path.join(dir, 'src', 'state-machine.ts');
+    fs.writeFileSync(touched, fs.readFileSync(touched, 'utf8').replace('OrderStateMachine', 'RenamedWorkflowEngine'));
+
+    expect((await cg.indexAll()).success).toBe(true);
+    expect(cg.getSegmentMatches(['renamed', 'workflow']).map((m) => m.name)).toContain('RenamedWorkflowEngine');
+    expect(cg.getSegmentMatches(['checkout', 'service']).map((m) => m.name)).toContain('CheckoutService');
+    expect(cg.getSegmentMatches(['state', 'machine'])).toEqual([]);
+
+    await cg.sync();
+    expect(cg.getSegmentMatches(['checkout', 'service']).map((m) => m.name)).toContain('CheckoutService');
+  });
+
+  it('preserves existing symbol words when full indexing is cancelled', async () => {
+    const before = cg.getSegmentMatches(['state', 'machine']);
+    const controller = new AbortController();
+    controller.abort();
+
+    const result = await cg.indexAll({ signal: controller.signal });
+    expect(result.success).toBe(false);
+    expect(result.errors.some((error) => error.message === 'Aborted')).toBe(true);
+    expect(cg.getNodesByName('OrderStateMachine')).toHaveLength(1);
+    expect(cg.getSegmentMatches(['state', 'machine'])).toEqual(before);
+  });
+
   it('single rare word qualifies; ubiquitous and singleton words do not', () => {
     // "checkout" clusters (Service + Controller) — a concept this repo is about.
     expect(cg.getSegmentMatches(['checkout']).map((m) => m.name)).toContain('CheckoutService');

src/index.ts

@@ -575,10 +575,12 @@ export class CodeGraph {
         // marker behind, so `codegraph status` can tell a truncated index
         // from a completed one instead of silently serving partial results.
         try { this.queries.setMetadata('index_state', 'indexing'); } catch { /* metadata is advisory */ }
-        // Segment vocabulary starts empty and is repopulated by the node write
-        // path as every file (re-)indexes below — so a full index is also the
-        // orphan-cleanup pass for names deleted since the last one.
-        try { this.queries.clearNameSegmentVocab(); } catch { /* vocab is advisory — never fail an index over it */ }
+        // Fresh graphs populate the vocabulary through node writes. Existing
+        // graphs skip unchanged files at the store, so keep their vocabulary
+        // until this run succeeds and rebuild it from all surviving names below.
+        if (freshDb) {
+          try { this.queries.clearNameSegmentVocab(); } catch { /* vocab is advisory — never fail an index over it */ }
+        }
         // Bulk FTS mode for the mass-insert phase: drop the per-row FTS sync
         // triggers, rebuild nodes_fts once from the nodes table afterwards.
         // Crash inside the window is healed on the next DatabaseConnection.open.
@@ -785,6 +787,13 @@ export class CodeGraph {
           }
         } catch { /* metadata is advisory — never fail an index over it */ }
 
+        if (result.success && !freshDb) {
+          try {
+            this.queries.clearNameSegmentVocab();
+            await this.rebuildNameSegmentVocab();
+          } catch { /* vocab is advisory — never fail an index over it */ }
+        }
+
         // The time covers the whole run, like the totals above. The
         // orchestrator's covers extraction alone, and the summary printed it
         // as the run's, leaving out resolution and linking (#2334).