fix(typescript): avoid quadratic ts-morph re-parsing with consolidateTypeFiles

[devin-ai-integration]

Description

Linear ticket: N/A

Auth0's node-auth0 regeneration (3,758 IR types, consolidateTypeFiles: true) has hit the 10‑minute remote task cap on every run since 2026-09-01. Production generator stdout stops at Generating /src/management/api/types/types.ts (2 MB / 40k lines) and emits nothing for the remaining ~9m54s; fiddle coordinator logs show run duration climbing with IR size (median 63% of the cap in June → 85% July → 92% Aug → 96% on the last success, Aug 27) at a fixed generator version, then a +15% IR jump tipped it over.

Root cause: generateConsolidatedTypeDeclarations did, per type,

const n = sourceFile.getStatements().length;   // walks the AST
generatedType.writeToFile(context);            // sourceFile.addStatements(...) -> full-text re-parse
sourceFile.insertStatements(n, newLine);       // another full-text re-parse

Every ts-morph manipulation re-parses the whole source file, so building one file from N types is O(N²) in file size. With thousands of types in a single file this dominates generation time.

Changes Made

• Collect generateStatements(context) for every type in the consolidated file (with the same blank-line separators) into one array and call sourceFile.addStatements(...) once.
• Changelog entry under generators/typescript/sdk/changes/unreleased/.
• Updated README.md generator (not applicable)

Testing

• pnpm seed test --generator ts-sdk --fixture exhaustive --outputFolder consolidate-type-files --skip-scripts passes; snapshot output unchanged.
• Manual benchmarks via seed run --generator ts-sdk --local --skip-scripts with consolidateTypeFiles: true, main vs this branch. diff -r of the full generated output (924 files) is empty in every case:

Input	main	this PR
Auth0 Management API spec + their generators.yml (3,758 IR types) — seed "Generation Time", same machine	6m 31.8s	1m 11.2s (~10s of which is codegen; rest is yarn lockfile)
Same, pure codegen measured separately	2m 49s	9.6s
Synthetic Fern definition, 3,000 object types	3m 0.7s	17.9s

With the fix, consolidateTypeFiles: true on the Auth0 spec is faster than the consolidateTypeFiles: false workaround (~47s).

Link to Devin session: https://app.devin.ai/sessions/22b0c13b5d0948a593e3b0e5444a48b7
Open in Devin Desktop: https://app.devin.ai/desktop/session/22b0c13b5d0948a593e3b0e5444a48b7?variant=devin

---

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment, CI, and merge conflict monitoring

[nitpickybot]

AI Review Summary

Solid perf fix: batching statements into a single addStatements call removes the quadratic re-parse. Main concern is a behavioral difference between writeToFile and generateStatements (side effects / import registration) and the empty-statements edge case.

• 🟡 2 warning(s)

---

To request another review, comment /ai-review on this pull request.

[nitpickybot]

🟡 warning

addStatements with an empty array still triggers a manipulation in some ts-morph versions; more importantly, guard against no-op work when typeDeclarations is empty:

Suggested change

	context.sourceFile.addStatements(statements);
	if (statements.length > 0) {
	context.sourceFile.addStatements(statements);
	}

[nitpickybot]

🟡 warning

Worth double-checking that writeToFile did nothing beyond sourceFile.addStatements(generateStatements(context)) for every GeneratedType implementation (alias, enum, union, undiscriminated union, object). If any override does extra work (e.g. adding imports, exports, or module declarations), that behavior is now silently dropped. The exhaustive fixture snapshot passing is good evidence but may not cover all variants (e.g. noOptionalProperties, retainOriginalCasing paths).

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no bugs or issues to report.

[github-actions]

SDK Generation Benchmark Results

Comparing PR branch against median of 5 nightly run(s) on main (latest: 2026-09-02T04:06:59Z).

Full benchmark table (click to expand)

Generator	Spec	main (generator)	main (E2E)	PR (generator)	Delta
ts-sdk	square	178s (n=5)	187s (n=5)	157s	-21s (-11.8%)

main (generator): generator-only time via --skip-scripts (includes Docker image build, container startup, IR parsing, and code generation — this is the same Docker-based flow customers use via fern generate). main (E2E): full customer-observable time including build/test scripts (nightly baseline, informational). Delta is computed against generator-only baseline.
⚠️ = generation exited with a non-zero exit code (timing may not reflect a successful run).
Baseline from nightly runs on main (latest: 2026-09-02T04:06:59Z). Trigger benchmark-baseline to refresh.
Last updated: 2026-09-02 22:26 UTC