fix(file-parsers): drain queued jobs on worker start failure, ship bundle under dist

Author: waleedlatif1

apps/sim/lib/file-parsers/xlsx-worker-entry.ts

@@ -1,6 +1,6 @@
 /**
  * Worker-thread entry for spreadsheet parsing, bundled to
- * `lib/file-parsers/dist/xlsx-worker.mjs` by `bun run build:xlsx-worker` and
+ * `apps/sim/dist/xlsx-worker.mjs` by `bun run build:xlsx-worker` and
  * spawned by `xlsx-worker-pool.ts`. It runs the same {@link XlsxParser} the
  * main thread falls back to, so the two paths cannot drift.
  */

apps/sim/lib/file-parsers/xlsx-worker-pool.test.ts

@@ -6,7 +6,8 @@
  * 1. Worker output diverges from in-process parsing (preview, complete).
  * 2. A typed parser error loses its class, `code`, or message crossing back.
  * 3. A missing bundle breaks parsing instead of falling back in-process.
- * 4. An unloadable bundle (throws at load) breaks parsing instead of falling back.
+ * 4. An unloadable bundle (throws at load, or the Worker constructor itself
+ *    throws) breaks parsing or strands queued jobs instead of falling back.
  * 5. A worker crash mid-parse loses the job instead of re-running it in-process.
  * 6. An abort during a worker parse hangs or leaves the worker occupying its slot.
  * 7. An abort while queued still dispatches the job.
@@ -151,6 +152,19 @@ describe('XlsxWorkerPool fallback', () => {
     await expect(pool.parse(input, {}, inProcess(input))).resolves.toEqual(expected)
   })
 
+  it('parses every queued job in-process when the Worker constructor throws', async () => {
+    const input = workbook()
+    const expected = await parser.parseBuffer(input)
+    const pool = new XlsxWorkerPool({ workerPath: 'not-a-resolvable-worker-path.mjs', size: 1 })
+
+    const results = await Promise.all([
+      pool.parse(input, {}, inProcess(input)),
+      pool.parse(input, {}, inProcess(input)),
+    ])
+
+    expect(results).toEqual([expected, expected])
+  })
+
   it('re-runs a job in-process when the worker dies mid-parse', async () => {
     const input = workbook()
     const expected = await parser.parseBuffer(input)

apps/sim/lib/file-parsers/xlsx-worker-pool.ts

@@ -28,7 +28,7 @@ const XLSX_WORKER_IDLE_TIMEOUT_MS = 30_000
 /** Bounds worker startup only; a parse itself is bounded by the caller's signal. */
 const XLSX_WORKER_START_TIMEOUT_MS = 10_000
 
-const WORKER_BUNDLE_RELATIVE_PATH = path.join('lib', 'file-parsers', 'dist', 'xlsx-worker.mjs')
+const WORKER_BUNDLE_RELATIVE_PATH = path.join('dist', 'xlsx-worker.mjs')
 
 /**
  * Where `bun run build:xlsx-worker` writes the bundle, as seen from the two
@@ -257,6 +257,7 @@ export class XlsxWorkerPool {
       this.retire(slot)
       if (job) this.runInProcess(job)
     }
+    for (const job of this.queue.splice(0)) this.runInProcess(job)
   }
 
   private runInProcess(job: Job): void {

apps/sim/lib/file-parsers/xlsx-worker-protocol.ts

@@ -11,7 +11,7 @@ import type { FileParseOptions, FileParseResult } from '@/lib/file-parsers/types
  * throw is encoded by kind and rebuilt as the same class on the main side.
  */
 
-export type XlsxWorkerParseOptions = Omit<FileParseOptions, 'signal'>
+type XlsxWorkerParseOptions = Omit<FileParseOptions, 'signal'>
 
 export interface XlsxWorkerParseRequest {
   type: 'parse'

apps/sim/next.config.ts

@@ -163,7 +163,7 @@ const nextConfig: NextConfig = {
      */
     '/*': [
       './lib/execution/sandbox/bundles/*.cjs',
-      './lib/file-parsers/dist/xlsx-worker.mjs',
+      './dist/xlsx-worker.mjs',
       './node_modules/ws/**/*',
       '../../packages/sim-cli/dist/runtime.js',
     ],

apps/sim/package.json

@@ -19,7 +19,7 @@
     "load:workflow:isolation": "BASE_URL=${BASE_URL:-http://localhost:3000} ISOLATION_DURATION=${ISOLATION_DURATION:-30} TOTAL_RATE=${TOTAL_RATE:-9} WORKSPACE_A_WEIGHT=${WORKSPACE_A_WEIGHT:-8} WORKSPACE_B_WEIGHT=${WORKSPACE_B_WEIGHT:-1} artillery run scripts/load/workflow-isolation.yml",
     "build": "bun run build:cli-runtime && bun run build:sandbox-bundles && bun run build:xlsx-worker && NODE_OPTIONS='--max-old-space-size=8192' next build",
     "build:sandbox-bundles": "bun run ./lib/execution/sandbox/bundles/build.ts",
-    "build:xlsx-worker": "bun build ./lib/file-parsers/xlsx-worker-entry.ts --target=node --format=esm --packages=bundle --outfile=./lib/file-parsers/dist/xlsx-worker.mjs",
+    "build:xlsx-worker": "bun build ./lib/file-parsers/xlsx-worker-entry.ts --target=node --format=esm --packages=bundle --outfile=./dist/xlsx-worker.mjs",
     "start": "next start",
     "test": "vitest run",
     "test:scim:e2e": "bun run scripts/test-scim-e2e.ts",

docker/app.Dockerfile

@@ -213,10 +213,11 @@ COPY --from=deps --chown=nextjs:nodejs /app/node_modules/@napi-rs ./node_modules
 # Copy the isolated-vm worker script
 COPY --from=builder --chown=nextjs:nodejs /app/apps/sim/lib/execution/isolated-vm-worker.cjs ./apps/sim/lib/execution/isolated-vm-worker.cjs
 
-# Spreadsheet parse worker bundle (build:xlsx-worker, run by `bun run build`).
+# Spreadsheet parse worker bundle (build:xlsx-worker, run by `bun run build`;
+# under the build task's declared `dist/**` output so a Turbo cache hit restores it).
 # Without it parsing falls back in-process, so a missing file fails the image
 # build here instead of silently putting XLSX parsing back on the event loop.
-COPY --from=builder --chown=nextjs:nodejs /app/apps/sim/lib/file-parsers/dist/xlsx-worker.mjs ./apps/sim/lib/file-parsers/dist/xlsx-worker.mjs
+COPY --from=builder --chown=nextjs:nodejs /app/apps/sim/dist/xlsx-worker.mjs ./apps/sim/dist/xlsx-worker.mjs
 
 # Copy the pre-built sandbox library bundles (pptxgenjs, docx, pdf-lib) that
 # run inside the V8 isolate. Committed into the repo; see