@optave/codegraph 3.10.0 → 3.11.1

package/src/domain/graph/builder/pipeline.ts

@@ -8,50 +8,24 @@ import fs from 'node:fs';
 import path from 'node:path';
 import { performance } from 'node:perf_hooks';
 import {
-  acquireAdvisoryLock,
   closeDb,
   closeDbPair,
   getBuildMeta,
   initSchema,
   MIGRATIONS,
   openDb,
-  releaseAdvisoryLock,
-  setBuildMeta,
 } from '../../../db/index.js';
 import { detectWorkspaces, loadConfig } from '../../../infrastructure/config.js';
 import { debug, info, warn } from '../../../infrastructure/logger.js';
 import { loadNative } from '../../../infrastructure/native.js';
-import { semverCompare } from '../../../infrastructure/update-check.js';
-import { normalizePath } from '../../../shared/constants.js';
 import { toErrorMessage } from '../../../shared/errors.js';
 import { CODEGRAPH_VERSION } from '../../../shared/version.js';
-import type {
-  BetterSqlite3Database,
-  BuildGraphOpts,
-  BuildResult,
-  Definition,
-  ExtractorOutput,
-  SqliteStatement,
-} from '../../../types.js';
-import {
-  classifyNativeDrops,
-  formatDropExtensionSummary,
-  getActiveEngine,
-  getInstalledWasmExtensions,
-  parseFilesWasmForBackfill,
-} from '../../parser.js';
+import type { BuildGraphOpts, BuildResult } from '../../../types.js';
+import { getActiveEngine } from '../../parser.js';
 import { writeJournalHeader } from '../journal.js';
 import { setWorkspaces } from '../resolve.js';
 import { PipelineContext } from './context.js';
-import {
-  batchInsertNodes,
-  collectFiles as collectFilesUtil,
-  fileHash,
-  fileStat,
-  loadPathAliases,
-  readFileSafe,
-} from './helpers.js';
-import { NativeDbProxy } from './native-db-proxy.js';
+import { loadPathAliases } from './helpers.js';
 import { buildEdges } from './stages/build-edges.js';
 import { buildStructure } from './stages/build-structure.js';
 // Pipeline stages
@@ -59,10 +33,24 @@ import { collectFiles } from './stages/collect-files.js';
 import { detectChanges, detectNoChanges } from './stages/detect-changes.js';
 import { finalize } from './stages/finalize.js';
 import { insertNodes } from './stages/insert-nodes.js';
+import {
+  closeNativeDb,
+  refreshJsDb,
+  reopenNativeDb,
+  suspendNativeDb,
+} from './stages/native-db-lifecycle.js';
+import { tryNativeOrchestrator } from './stages/native-orchestrator.js';
 import { parseFiles } from './stages/parse-files.js';
 import { resolveImports } from './stages/resolve-imports.js';
 import { runAnalyses } from './stages/run-analyses.js';
 
+// Re-export computeWasmOnlyStaleFiles for backward compatibility with tests
+// that import from this module path (#1073 unit tests).
+export {
+  computeWasmOnlyStaleFiles,
+  type WasmOnlyStaleFilesInput,
+} from './stages/native-orchestrator.js';
+
 // ── Setup helpers ───────────────────────────────────────────────────────
 
 function initializeEngine(ctx: PipelineContext): void {
@@ -165,7 +153,9 @@ function loadAliases(ctx: PipelineContext): void {
 
 function setupPipeline(ctx: PipelineContext): void {
   ctx.rootDir = path.resolve(ctx.rootDir);
-  ctx.dbPath = path.join(ctx.rootDir, '.codegraph', 'graph.db');
+  ctx.dbPath = ctx.opts.dbPath
+    ? path.resolve(ctx.opts.dbPath)
+    : path.join(ctx.rootDir, '.codegraph', 'graph.db');
 
   // Detect whether native engine is available.
   const enginePref = ctx.opts.engine || 'auto';
@@ -183,6 +173,16 @@ function setupPipeline(ctx: PipelineContext): void {
   initSchema(ctx.db);
 
   ctx.config = loadConfig(ctx.rootDir);
+  // Merge caller-supplied excludes on top of the file-config excludes so
+  // programmatic callers (e.g. benchmark scripts) can extend exclusion
+  // without mutating .codegraphrc.json. Native orchestrator picks this up
+  // automatically — it reads exclude off the serialized ctx.config below.
+  if (ctx.opts.exclude?.length) {
+    ctx.config = {
+      ...ctx.config,
+      exclude: [...(ctx.config.exclude ?? []), ...ctx.opts.exclude],
+    };
+  }
   ctx.incremental =
     ctx.opts.incremental !== false && ctx.config.build && ctx.config.build.incremental !== false;
 
@@ -223,761 +223,8 @@ function formatTimingResult(ctx: PipelineContext): BuildResult {
   };
 }
 
-// ── NativeDb lifecycle helpers ──────────────────────────────────────────
-
-/** Checkpoint WAL through rusqlite and close the native connection. */
-function closeNativeDb(ctx: PipelineContext, label: string): void {
-  if (!ctx.nativeDb) return;
-  try {
-    ctx.nativeDb.exec('PRAGMA wal_checkpoint(TRUNCATE)');
-  } catch (e) {
-    debug(`${label} WAL checkpoint failed: ${toErrorMessage(e)}`);
-  }
-  try {
-    ctx.nativeDb.close();
-  } catch (e) {
-    debug(`${label} nativeDb close failed: ${toErrorMessage(e)}`);
-  }
-  ctx.nativeDb = undefined;
-}
-
-/** Try to reopen the native connection for a given pipeline phase. */
-function reopenNativeDb(ctx: PipelineContext, label: string): void {
-  if ((ctx.opts.engine ?? 'auto') === 'wasm') return;
-  const native = loadNative();
-  if (!native?.NativeDatabase) return;
-  try {
-    ctx.nativeDb = native.NativeDatabase.openReadWrite(ctx.dbPath);
-  } catch (e) {
-    debug(`reopen nativeDb for ${label} failed: ${toErrorMessage(e)}`);
-    ctx.nativeDb = undefined;
-  }
-}
-
-/** Close nativeDb and clear stale references in engineOpts. */
-function suspendNativeDb(ctx: PipelineContext, label: string): void {
-  closeNativeDb(ctx, label);
-  if (ctx.engineOpts?.nativeDb) {
-    ctx.engineOpts.nativeDb = undefined;
-  }
-}
-
-/**
- * After native writes, reopen the JS db connection to get a fresh page cache.
- * Rusqlite WAL truncation invalidates better-sqlite3's internal WAL index,
- * causing SQLITE_CORRUPT on the next read (#715, #736).
- */
-function refreshJsDb(ctx: PipelineContext): void {
-  try {
-    ctx.db.close();
-  } catch (e) {
-    debug(`refreshJsDb close failed: ${toErrorMessage(e)}`);
-  }
-  ctx.db = openDb(ctx.dbPath);
-}
-
-// ── Native orchestrator types ──────────────────────────────────────────
-
-interface NativeOrchestratorResult {
-  phases: Record<string, number>;
-  earlyExit?: boolean;
-  nodeCount?: number;
-  edgeCount?: number;
-  fileCount?: number;
-  changedFiles?: string[];
-  changedCount?: number;
-  removedCount?: number;
-  isFullBuild?: boolean;
-  /** Whether the Rust pipeline handled the structure phase (small-incremental fast path). */
-  structureHandled?: boolean;
-  /** Whether the Rust pipeline wrote AST/complexity/CFG/dataflow to DB. */
-  analysisComplete?: boolean;
-}
-
-// ── Native orchestrator helpers ───────────────────────────────────────
-
-/** Determine whether the native orchestrator should be skipped. Returns a reason string, or null if it should run. */
-function shouldSkipNativeOrchestrator(ctx: PipelineContext): string | null {
-  if (ctx.forceFullRebuild) return 'forceFullRebuild';
-  // v3.9.0 addon had buggy incremental purge (wrong SQL on analysis tables,
-  // scoped removal over-detection). Fixed in v3.9.1 by PR #865. Gate on
-  // < 3.9.1 so v3.9.1+ uses the fast Rust orchestrator path.
-  const orchestratorBuggy = !!ctx.engineVersion && semverCompare(ctx.engineVersion, '3.9.1') < 0;
-  if (orchestratorBuggy) return `buggy addon ${ctx.engineVersion}`;
-  if (ctx.engineName !== 'native') return `engine=${ctx.engineName}`;
-  return null;
-}
-
-/** Checkpoint WAL through rusqlite, close nativeDb, and reopen better-sqlite3.
- *  Returns false if the DB reopen fails (caller should return partial result). */
-function handoffWalAfterNativeBuild(ctx: PipelineContext): boolean {
-  closeNativeDb(ctx, 'post-native-build');
-  try {
-    ctx.db.close();
-  } catch (e) {
-    debug(`handoffWal JS db close failed: ${toErrorMessage(e)}`);
-  }
-  try {
-    ctx.db = openDb(ctx.dbPath);
-    return true;
-  } catch (reopenErr) {
-    warn(`Failed to reopen DB after native build: ${(reopenErr as Error).message}`);
-    return false;
-  }
-}
-
-/**
- * Reconstruct fileSymbols from the DB after a native orchestrator build.
- * When `scopeFiles` is provided, only loads those files (for analysis-only).
- * When omitted, loads all files (needed for structure rebuilds).
- */
-function reconstructFileSymbolsFromDb(
-  ctx: PipelineContext,
-  scopeFiles?: string[],
-): Map<string, ExtractorOutput> {
-  let query =
-    'SELECT file, name, kind, line, end_line as endLine FROM nodes WHERE file IS NOT NULL';
-  const params: string[] = [];
-  if (scopeFiles && scopeFiles.length > 0) {
-    const placeholders = scopeFiles.map(() => '?').join(',');
-    query += ` AND file IN (${placeholders})`;
-    params.push(...scopeFiles);
-  }
-  query += ' ORDER BY file, line';
-
-  const rows = ctx.db.prepare(query).all(...params) as {
-    file: string;
-    name: string;
-    kind: string;
-    line: number;
-    endLine: number | null;
-  }[];
-
-  const fileSymbols = new Map<string, ExtractorOutput>();
-  for (const row of rows) {
-    let entry = fileSymbols.get(row.file);
-    if (!entry) {
-      entry = {
-        definitions: [],
-        calls: [],
-        imports: [],
-        classes: [],
-        exports: [],
-        typeMap: new Map(),
-      };
-      fileSymbols.set(row.file, entry);
-    }
-    entry.definitions.push({
-      name: row.name,
-      kind: row.kind as Definition['kind'],
-      line: row.line,
-      endLine: row.endLine ?? undefined,
-    });
-  }
-
-  // Populate import/export counts from DB edges so buildStructure
-  // computes correct import_count/export_count in node_metrics.
-  // The extractor arrays aren't persisted to the DB, so we derive
-  // counts from edge data instead (#804).
-  const importCountRows = ctx.db
-    .prepare(
-      `SELECT n.file, COUNT(*) AS cnt
-       FROM edges e JOIN nodes n ON e.source_id = n.id
-       WHERE e.kind IN ('imports', 'imports-type', 'dynamic-imports')
-         AND n.file IS NOT NULL
-       GROUP BY n.file`,
-    )
-    .all() as { file: string; cnt: number }[];
-  for (const row of importCountRows) {
-    const entry = fileSymbols.get(row.file);
-    if (entry) entry.imports = new Array(row.cnt) as ExtractorOutput['imports'];
-  }
-
-  const exportCountRows = ctx.db
-    .prepare(
-      `SELECT n_tgt.file, COUNT(DISTINCT n_tgt.id) AS cnt
-       FROM edges e
-       JOIN nodes n_tgt ON e.target_id = n_tgt.id
-       JOIN nodes n_src ON e.source_id = n_src.id
-       WHERE e.kind IN ('imports', 'imports-type', 'reexports')
-         AND n_tgt.file IS NOT NULL
-         AND n_src.file != n_tgt.file
-       GROUP BY n_tgt.file`,
-    )
-    .all() as { file: string; cnt: number }[];
-  for (const row of exportCountRows) {
-    const entry = fileSymbols.get(row.file);
-    if (entry) entry.exports = new Array(row.cnt) as ExtractorOutput['exports'];
-  }
-
-  return fileSymbols;
-}
-
-/**
- * Run JS buildStructure() after native orchestrator to fill directory nodes + contains edges.
- * For full builds, passes changedFiles=null (full rebuild).
- * For incremental builds, passes the changed file list to scope the update.
- */
-async function runPostNativeStructure(
-  ctx: PipelineContext,
-  allFileSymbols: Map<string, ExtractorOutput>,
-  isFullBuild: boolean,
-  changedFiles: string[] | undefined,
-): Promise<number> {
-  const structureStart = performance.now();
-  try {
-    const directories = new Set<string>();
-    for (const relPath of allFileSymbols.keys()) {
-      const parts = relPath.split('/');
-      for (let i = 1; i < parts.length; i++) {
-        directories.add(parts.slice(0, i).join('/'));
-      }
-    }
-
-    const lineCountMap = new Map<string, number>();
-    const cachedLineCounts = ctx.db
-      .prepare(
-        `SELECT n.name AS file, m.line_count
-         FROM node_metrics m JOIN nodes n ON m.node_id = n.id
-         WHERE n.kind = 'file'`,
-      )
-      .all() as Array<{ file: string; line_count: number }>;
-    for (const row of cachedLineCounts) {
-      lineCountMap.set(row.file, row.line_count);
-    }
-
-    // Full builds need null (rebuild everything). Incremental builds pass the
-    // changed file list so buildStructure only updates those files' metrics
-    // and contains edges — matching the JS pipeline's medium-incremental path.
-    const changedFilePaths = isFullBuild || !changedFiles?.length ? null : changedFiles;
-    const { buildStructure: buildStructureFn } = (await import(
-      '../../../features/structure.js'
-    )) as {
-      buildStructure: (
-        db: typeof ctx.db,
-        fileSymbols: Map<string, ExtractorOutput>,
-        rootDir: string,
-        lineCountMap: Map<string, number>,
-        directories: Set<string>,
-        changedFiles: string[] | null,
-      ) => void;
-    };
-    buildStructureFn(
-      ctx.db,
-      allFileSymbols,
-      ctx.rootDir,
-      lineCountMap,
-      directories,
-      changedFilePaths,
-    );
-    debug(
-      `Structure phase completed after native orchestrator${changedFilePaths ? ` (${changedFilePaths.length} files)` : ' (full)'}`,
-    );
-  } catch (err) {
-    warn(`Structure phase failed after native build: ${toErrorMessage(err)}`);
-  }
-  return performance.now() - structureStart;
-}
-
-/**
- * JS fallback for AST/complexity/CFG/dataflow analysis after native orchestrator.
- * Used when the Rust addon doesn't include analysis persistence (older addon
- * version) or when analysis failed on the Rust side.
- */
-async function runPostNativeAnalysis(
-  ctx: PipelineContext,
-  allFileSymbols: Map<string, ExtractorOutput>,
-  changedFiles: string[] | undefined,
-): Promise<{ astMs: number; complexityMs: number; cfgMs: number; dataflowMs: number }> {
-  const timing = { astMs: 0, complexityMs: 0, cfgMs: 0, dataflowMs: 0 };
-
-  // Scope analysis fileSymbols to changed files only
-  let analysisFileSymbols: Map<string, ExtractorOutput>;
-  if (changedFiles && changedFiles.length > 0) {
-    analysisFileSymbols = new Map();
-    for (const f of changedFiles) {
-      const entry = allFileSymbols.get(f);
-      if (entry) analysisFileSymbols.set(f, entry);
-    }
-  } else {
-    analysisFileSymbols = allFileSymbols;
-  }
-
-  // Reopen nativeDb for analysis features (suspend/resume WAL pattern).
-  const native = loadNative();
-  if (native?.NativeDatabase) {
-    try {
-      ctx.nativeDb = native.NativeDatabase.openReadWrite(ctx.dbPath);
-      if (ctx.engineOpts) ctx.engineOpts.nativeDb = ctx.nativeDb;
-    } catch {
-      ctx.nativeDb = undefined;
-      if (ctx.engineOpts) ctx.engineOpts.nativeDb = undefined;
-    }
-  }
-
-  // Flush JS WAL pages once so Rust can see them, then no-op callbacks.
-  // Previously each feature called wal_checkpoint(TRUNCATE) individually
-  // (~68ms each × 3-4 features). One FULL checkpoint suffices.
-  if (ctx.nativeDb && ctx.engineOpts) {
-    ctx.db.pragma('wal_checkpoint(FULL)');
-    ctx.engineOpts.suspendJsDb = () => {};
-    ctx.engineOpts.resumeJsDb = () => {};
-  }
-
-  try {
-    const { runAnalyses: runAnalysesFn } = (await import('../../../ast-analysis/engine.js')) as {
-      runAnalyses: (
-        db: BetterSqlite3Database,
-        fileSymbols: Map<string, ExtractorOutput>,
-        rootDir: string,
-        opts: Record<string, unknown>,
-        engineOpts?: Record<string, unknown>,
-      ) => Promise<{ astMs?: number; complexityMs?: number; cfgMs?: number; dataflowMs?: number }>;
-    };
-    const result = await runAnalysesFn(
-      ctx.db,
-      analysisFileSymbols,
-      ctx.rootDir,
-      ctx.opts as Record<string, unknown>,
-      ctx.engineOpts as unknown as Record<string, unknown> | undefined,
-    );
-    timing.astMs = result.astMs ?? 0;
-    timing.complexityMs = result.complexityMs ?? 0;
-    timing.cfgMs = result.cfgMs ?? 0;
-    timing.dataflowMs = result.dataflowMs ?? 0;
-  } catch (err) {
-    warn(`Analysis phases failed after native build: ${toErrorMessage(err)}`);
-  }
-
-  // Close nativeDb after analyses — TRUNCATE checkpoint flushes all Rust
-  // WAL writes so JS and external readers can see them. Runs once after
-  // all analysis features complete (not per-feature).
-  if (ctx.nativeDb) {
-    try {
-      ctx.nativeDb.exec('PRAGMA wal_checkpoint(TRUNCATE)');
-    } catch {
-      /* ignore checkpoint errors */
-    }
-    try {
-      ctx.nativeDb.close();
-    } catch {
-      /* ignore close errors */
-    }
-    ctx.nativeDb = undefined;
-    if (ctx.engineOpts) {
-      ctx.engineOpts.nativeDb = undefined;
-      ctx.engineOpts.suspendJsDb = undefined;
-      ctx.engineOpts.resumeJsDb = undefined;
-    }
-  }
-
-  return timing;
-}
-
-/** Format timing result from native orchestrator phases + JS post-processing. */
-function formatNativeTimingResult(
-  p: Record<string, number>,
-  structurePatchMs: number,
-  analysisTiming: { astMs: number; complexityMs: number; cfgMs: number; dataflowMs: number },
-): BuildResult {
-  return {
-    phases: {
-      setupMs: +(p.setupMs ?? 0).toFixed(1),
-      collectMs: +(p.collectMs ?? 0).toFixed(1),
-      detectMs: +(p.detectMs ?? 0).toFixed(1),
-      parseMs: +(p.parseMs ?? 0).toFixed(1),
-      insertMs: +(p.insertMs ?? 0).toFixed(1),
-      resolveMs: +(p.resolveMs ?? 0).toFixed(1),
-      edgesMs: +(p.edgesMs ?? 0).toFixed(1),
-      structureMs: +((p.structureMs ?? 0) + structurePatchMs).toFixed(1),
-      rolesMs: +(p.rolesMs ?? 0).toFixed(1),
-      astMs: +(analysisTiming.astMs ?? 0).toFixed(1),
-      complexityMs: +(analysisTiming.complexityMs ?? 0).toFixed(1),
-      cfgMs: +(analysisTiming.cfgMs ?? 0).toFixed(1),
-      dataflowMs: +(analysisTiming.dataflowMs ?? 0).toFixed(1),
-      finalizeMs: +(p.finalizeMs ?? 0).toFixed(1),
-    },
-  };
-}
-
-/** Try the native build orchestrator. Returns a BuildResult on success, undefined to fall through to JS pipeline. */
-async function tryNativeOrchestrator(
-  ctx: PipelineContext,
-): Promise<BuildResult | undefined | 'early-exit'> {
-  const skipReason = shouldSkipNativeOrchestrator(ctx);
-  if (skipReason) {
-    debug(`Skipping native orchestrator: ${skipReason}`);
-    return undefined;
-  }
-
-  // Open NativeDatabase on demand — deferred from setupPipeline to skip the
-  // ~60ms cost on no-op/early-exit builds. Close the better-sqlite3 connection
-  // first to avoid dual-connection WAL corruption.
-  if (!ctx.nativeDb && ctx.nativeAvailable) {
-    const native = loadNative();
-    if (native?.NativeDatabase) {
-      try {
-        // Close better-sqlite3 before opening rusqlite to avoid WAL conflicts.
-        // Uses raw close() instead of closeDb() intentionally — the advisory lock
-        // is kept and transferred to the NativeDbProxy below, not released here.
-        ctx.db.close();
-        acquireAdvisoryLock(ctx.dbPath);
-        ctx.nativeDb = native.NativeDatabase.openReadWrite(ctx.dbPath);
-        ctx.nativeDb.initSchema();
-        // Replace ctx.db with a NativeDbProxy so post-native JS fallback
-        // (structure, analysis) can use it without reopening better-sqlite3.
-        const proxy = new NativeDbProxy(ctx.nativeDb);
-        proxy.__lockPath = `${ctx.dbPath}.lock`;
-        ctx.db = proxy as unknown as typeof ctx.db;
-        ctx.nativeFirstProxy = true;
-      } catch (err) {
-        warn(`NativeDatabase setup failed, falling back to JS: ${toErrorMessage(err)}`);
-        try {
-          ctx.nativeDb?.close();
-        } catch (e) {
-          debug(`tryNativeOrchestrator: close failed during fallback: ${toErrorMessage(e)}`);
-        }
-        ctx.nativeDb = undefined;
-        ctx.nativeFirstProxy = false; // defensive: reset in case future refactors move the assignment above throwing lines
-        releaseAdvisoryLock(`${ctx.dbPath}.lock`);
-        // Reopen better-sqlite3 for JS pipeline fallback
-        ctx.db = openDb(ctx.dbPath);
-      }
-    }
-  }
-
-  if (!ctx.nativeDb?.buildGraph) return undefined;
-
-  const resultJson = ctx.nativeDb.buildGraph(
-    ctx.rootDir,
-    JSON.stringify(ctx.config),
-    JSON.stringify(ctx.aliases),
-    JSON.stringify(ctx.opts),
-  );
-  const result = JSON.parse(resultJson) as NativeOrchestratorResult;
-
-  if (result.earlyExit) {
-    info('No changes detected');
-    closeDbPair({ db: ctx.db, nativeDb: ctx.nativeDb });
-    return 'early-exit';
-  }
-
-  // Log incremental status to match JS pipeline output
-  const changed = result.changedCount ?? 0;
-  const removed = result.removedCount ?? 0;
-  if (!result.isFullBuild && (changed > 0 || removed > 0)) {
-    info(`Incremental: ${changed} changed, ${removed} removed`);
-  }
-
-  const p = result.phases;
-
-  // Sync build_meta so JS-side version/engine checks work on next build.
-  // Use the binary's CARGO_PKG_VERSION (ctx.nativeBinaryVersion), not the
-  // platform package.json version (ctx.engineVersion). The Rust side's
-  // check_version_mismatch compares against CARGO_PKG_VERSION; writing
-  // the package.json value would create a permanent mismatch whenever
-  // the binary and platform package.json diverge — e.g., CI hot-swap
-  // via ci-install-native.mjs (#1066) — forcing every subsequent build
-  // to be a full rebuild.
-  //
-  // When the native addon doesn't expose engineVersion() (older addon),
-  // fall back to CODEGRAPH_VERSION — same fallback used by both
-  // checkEngineSchemaMismatch (read path) and persistBuildMetadata
-  // (the JS-pipeline write path in finalize.ts). Using ctx.engineVersion
-  // here would re-introduce the asymmetry this PR fixes for that case.
-  const nativeVersionForMeta = ctx.nativeBinaryVersion || CODEGRAPH_VERSION;
-  setBuildMeta(ctx.db, {
-    engine: ctx.engineName,
-    engine_version: nativeVersionForMeta,
-    codegraph_version: nativeVersionForMeta,
-    schema_version: String(ctx.schemaVersion),
-    built_at: new Date().toISOString(),
-  });
-
-  info(
-    `Native build orchestrator completed: ${result.nodeCount ?? 0} nodes, ${result.edgeCount ?? 0} edges, ${result.fileCount ?? 0} files`,
-  );
-
-  // ── Post-native structure + analysis ──────────────────────────────
-  let analysisTiming = {
-    astMs: +(p.astMs ?? 0),
-    complexityMs: +(p.complexityMs ?? 0),
-    cfgMs: +(p.cfgMs ?? 0),
-    dataflowMs: +(p.dataflowMs ?? 0),
-  };
-  let structurePatchMs = 0;
-  // Skip JS structure when the Rust pipeline's small-incremental fast path
-  // already handled it. For full builds and large incrementals where Rust
-  // skipped structure, we must run the JS fallback.
-  const needsStructure = !result.structureHandled;
-  // When the Rust addon doesn't include analysis persistence (older addon
-  // version or analysis failed), fall back to JS-side analysis.
-  const needsAnalysisFallback =
-    !result.analysisComplete &&
-    (ctx.opts.ast !== false ||
-      ctx.opts.complexity !== false ||
-      ctx.opts.cfg !== false ||
-      ctx.opts.dataflow !== false);
-
-  if (needsStructure || needsAnalysisFallback) {
-    // When analysis fallback is needed, handoff to better-sqlite3 — the
-    // analysis engine uses the suspend/resume WAL pattern that requires a
-    // real better-sqlite3 connection, not the NativeDbProxy.
-    if (needsAnalysisFallback && ctx.nativeFirstProxy) {
-      closeNativeDb(ctx, 'pre-analysis-fallback');
-      ctx.db = openDb(ctx.dbPath);
-      ctx.nativeFirstProxy = false;
-    } else if (!ctx.nativeFirstProxy && !handoffWalAfterNativeBuild(ctx)) {
-      // DB reopen failed — return partial result
-      return formatNativeTimingResult(p, 0, analysisTiming);
-    }
-
-    const fileSymbols = reconstructFileSymbolsFromDb(ctx);
-
-    if (needsStructure) {
-      structurePatchMs = await runPostNativeStructure(
-        ctx,
-        fileSymbols,
-        !!result.isFullBuild,
-        result.changedFiles,
-      );
-    }
-
-    if (needsAnalysisFallback) {
-      analysisTiming = await runPostNativeAnalysis(ctx, fileSymbols, result.changedFiles);
-    }
-  }
-
-  // Engine parity: the native orchestrator silently drops files whose
-  // Rust extractor/grammar is missing or fails (e.g. HCL, Scala, Swift on
-  // stale native binaries). WASM handles those — backfill via WASM so both
-  // engines process the same file set (#967).
-  //
-  // Runs on full builds and on incrementals when the orchestrator reports
-  // any file activity (removals or changes). The orchestrator's
-  // `detect_removed_files` filter (#1070) skips files outside its narrower
-  // file_collector, so on a current binary a no-op rebuild reports
-  // `removedCount=0` and `changedCount=0`, making the backfill call pure
-  // overhead (fs walk + 2 DB queries + 48-file WASM re-parse). Legacy
-  // binaries lacking the filter still report `removedCount>0` and get the
-  // gap-repair behavior #1068 introduced. Triggering on `changedCount>0`
-  // narrows (but does not fully close) the gap where a brand-new
-  // unsupported-extension file is added on an otherwise-quiet incremental
-  // — see #1091 for the residual gap.
-  const removedCount = result.removedCount ?? 0;
-  const changedCount = result.changedCount ?? 0;
-  if (result.isFullBuild || removedCount > 0 || changedCount > 0) {
-    await backfillNativeDroppedFiles(ctx);
-  }
-
-  closeDbPair({ db: ctx.db, nativeDb: ctx.nativeDb });
-  return formatNativeTimingResult(p, structurePatchMs, analysisTiming);
-}
-
-/**
- * Backfill files that the native orchestrator silently dropped during parse.
- * Falls back to WASM + inserts file/symbol nodes so engine counts match (#967).
- */
-async function backfillNativeDroppedFiles(ctx: PipelineContext): Promise<void> {
-  // Compute the missing-file set FIRST, before any expensive DB handoff.
-  // NativeDbProxy supports .prepare().all(), so the upfront query works
-  // whether ctx.db is a proxy or a real better-sqlite3 connection. On
-  // incremental no-op rebuilds nothing is missing, so we want to early-return
-  // without paying the close-native / reopen-better-sqlite3 cost.
-  const collected = collectFilesUtil(ctx.rootDir, [], ctx.config, new Set<string>());
-  const expected = new Set(
-    collected.files.map((f) => normalizePath(path.relative(ctx.rootDir, f))),
-  );
-
-  const existingNodeRows = ctx.db
-    .prepare("SELECT DISTINCT file FROM nodes WHERE kind = 'file'")
-    .all() as Array<{ file: string }>;
-  const existingNodes = new Set(existingNodeRows.map((r) => r.file));
-
-  // Belt-and-suspenders: also check `file_hashes`. The fast-skip pre-flight
-  // (#1054) rejects on `file_hashes` gaps, and the two tables can diverge
-  // (e.g. a DB written by old code where `nodes` was populated but
-  // `file_hashes` was not). Treating "in nodes but not in file_hashes" as
-  // missing closes the gap so the backfill repairs the file_hashes row even
-  // when the node row already exists.
-  let existingHashes = new Set<string>();
-  try {
-    const existingHashRows = ctx.db
-      .prepare('SELECT DISTINCT file FROM file_hashes')
-      .all() as Array<{ file: string }>;
-    existingHashes = new Set(existingHashRows.map((r) => r.file));
-  } catch (e) {
-    // file_hashes table may not exist on legacy DBs; treat as fully missing
-    // so the backfill writes rows on the upsert path below.
-    debug(
-      `backfillNativeDroppedFiles: file_hashes read failed (table may not exist): ${toErrorMessage(e)}`,
-    );
-  }
-
-  // Restrict backfill to files with an installed WASM grammar. Extensions in
-  // LANGUAGE_REGISTRY without a shipped grammar file (e.g. groovy, erlang on
-  // minimal installs) can't be parsed by either engine, so they're not a
-  // native regression — excluding them keeps the warn count meaningful.
-  const installedExts = getInstalledWasmExtensions();
-  const missingRel: string[] = [];
-  const missingAbs: string[] = [];
-  for (const rel of expected) {
-    // A file is "missing" if it's absent from EITHER nodes OR file_hashes.
-    // Both must be present for fast-skip to work correctly.
-    if (existingNodes.has(rel) && existingHashes.has(rel)) continue;
-    const ext = path.extname(rel).toLowerCase();
-    if (!installedExts.has(ext)) continue;
-    missingRel.push(rel);
-    missingAbs.push(path.join(ctx.rootDir, rel));
-  }
-  if (missingAbs.length === 0) return;
-
-  // Now that we know there's work to do, hand off to better-sqlite3 (needed
-  // for the INSERT path below).
-  if (ctx.nativeFirstProxy) {
-    closeNativeDb(ctx, 'pre-parity-backfill');
-    ctx.db = openDb(ctx.dbPath);
-    ctx.nativeFirstProxy = false;
-  }
-
-  // Classify drops so users see per-extension reasons instead of just a count
-  // (#1011). `unsupported-by-native` is a legitimate parser limit (no Rust
-  // extractor); `native-extractor-failure` indicates a real native bug since
-  // the language IS supported by the addon yet the file was dropped anyway.
-  const { byReason, totals } = classifyNativeDrops(missingRel);
-  if (totals['unsupported-by-native'] > 0) {
-    info(
-      `Native orchestrator skipped ${totals['unsupported-by-native']} file(s) in languages without a Rust extractor; backfilling via WASM: ${formatDropExtensionSummary(byReason['unsupported-by-native'])}`,
-    );
-  }
-  if (totals['native-extractor-failure'] > 0) {
-    warn(
-      `Native orchestrator dropped ${totals['native-extractor-failure']} file(s) in natively-supported languages — likely a Rust extractor bug. Backfilling via WASM: ${formatDropExtensionSummary(byReason['native-extractor-failure'])}`,
-    );
-  }
-  const wasmResults = await parseFilesWasmForBackfill(missingAbs, ctx.rootDir);
-
-  const rows: unknown[][] = [];
-  const exportKeys: unknown[][] = [];
-  for (const [relPath, symbols] of wasmResults) {
-    // File row — mirrors insertDefinitionsAndExports: qualified_name is null.
-    rows.push([relPath, 'file', relPath, 0, null, null, null, null, null]);
-    for (const def of symbols.definitions ?? []) {
-      // Populate qualified_name/scope the same way the JS fallback does so
-      // downstream queries (cross-file references, "go to definition") find
-      // these symbols.
-      const dotIdx = def.name.lastIndexOf('.');
-      const scope = dotIdx !== -1 ? def.name.slice(0, dotIdx) : null;
-      rows.push([
-        def.name,
-        def.kind,
-        relPath,
-        def.line,
-        def.endLine ?? null,
-        null,
-        def.name,
-        scope,
-        def.visibility ?? null,
-      ]);
-    }
-    // Exports: insert the row (INSERT OR IGNORE — a matching definition row
-    // is a no-op) and queue a key for the second-pass exported=1 update, so
-    // queries filtering on exported=1 find backfilled symbols (#970).
-    for (const exp of symbols.exports ?? []) {
-      rows.push([exp.name, exp.kind, relPath, exp.line, null, null, exp.name, null, null]);
-      exportKeys.push([exp.name, exp.kind, relPath, exp.line]);
-    }
-  }
-  const db = ctx.db as unknown as BetterSqlite3Database;
-  batchInsertNodes(db, rows);
-
-  // Mark exported symbols in batches — mirrors insertDefinitionsAndExports.
-  if (exportKeys.length > 0) {
-    const EXPORT_CHUNK = 500;
-    const exportStmtCache = new Map<number, SqliteStatement>();
-    for (let i = 0; i < exportKeys.length; i += EXPORT_CHUNK) {
-      const end = Math.min(i + EXPORT_CHUNK, exportKeys.length);
-      const chunkSize = end - i;
-      let updateStmt = exportStmtCache.get(chunkSize);
-      if (!updateStmt) {
-        const conditions = Array.from(
-          { length: chunkSize },
-          () => '(name = ? AND kind = ? AND file = ? AND line = ?)',
-        ).join(' OR ');
-        updateStmt = db.prepare(`UPDATE nodes SET exported = 1 WHERE ${conditions}`);
-        exportStmtCache.set(chunkSize, updateStmt);
-      }
-      const vals: unknown[] = [];
-      for (let j = i; j < end; j++) {
-        const k = exportKeys[j] as unknown[];
-        vals.push(k[0], k[1], k[2], k[3]);
-      }
-      updateStmt.run(...vals);
-    }
-  }
-
-  // Persist file_hashes rows for every backfilled file. The Rust orchestrator
-  // only hashes files it parsed itself, so without this step files in
-  // optional-language extensions (e.g. .clj when no Rust extractor exists)
-  // would be missing from `file_hashes` — permanently breaking the JS-side
-  // fast-skip pre-flight (#1054), which rejects on `collected file missing
-  // from file_hashes` and forces every no-op rebuild back through the full
-  // ~2s native pipeline (#1068).
-  //
-  // Iterates `missingRel` (every collected file the Rust orchestrator
-  // dropped), not `wasmResults`, so files that produced zero symbols still
-  // get a row.
-  try {
-    const upsertHash = db.prepare(
-      'INSERT OR REPLACE INTO file_hashes (file, hash, mtime, size) VALUES (?, ?, ?, ?)',
-    );
-    const writeHashes = db.transaction(() => {
-      for (let i = 0; i < missingRel.length; i++) {
-        const relPath = missingRel[i];
-        const absPath = missingAbs[i];
-        if (!relPath || !absPath) continue;
-        let code: string | null;
-        try {
-          code = readFileSafe(absPath);
-        } catch (e) {
-          debug(`backfillNativeDroppedFiles: read failed for ${relPath}: ${toErrorMessage(e)}`);
-          continue;
-        }
-        if (code === null) continue;
-        const stat = fileStat(absPath);
-        const mtime = stat ? stat.mtime : 0;
-        const size = stat ? stat.size : 0;
-        upsertHash.run(relPath, fileHash(code), mtime, size);
-      }
-    });
-    writeHashes();
-  } catch (e) {
-    debug(
-      `backfillNativeDroppedFiles: file_hashes write failed (table may not exist): ${toErrorMessage(e)}`,
-    );
-  }
-
-  // Free WASM parse trees from the inline backfill path (#1058).
-  // `parseFilesWasmInline` sets `symbols._tree` (a live web-tree-sitter Tree
-  // backed by WASM linear memory) on every result, but these symbols are
-  // consumed locally for DB row construction and never added to
-  // `ctx.allSymbols`, so the finalize-stage `releaseWasmTrees` sweep never
-  // sees them. Without this, trees leak WASM memory until process exit —
-  // bounded per run but cumulative across in-process integration tests.
-  // Mirrors the cleanup discipline established for #931.
-  for (const [, symbols] of wasmResults) {
-    const tree = (symbols as { _tree?: { delete?: () => void } })._tree;
-    if (tree && typeof tree.delete === 'function') {
-      try {
-        tree.delete();
-      } catch {
-        /* ignore cleanup errors */
-      }
-    }
-    (symbols as { _tree?: unknown; _langId?: unknown })._tree = undefined;
-    (symbols as { _tree?: unknown; _langId?: unknown })._langId = undefined;
-  }
-}
+// Native db lifecycle and orchestrator helpers live in dedicated stage
+// modules — see `./stages/native-db-lifecycle.ts` and `./stages/native-orchestrator.ts`.
 
 // ── Pipeline stages execution ───────────────────────────────────────────