[johnnyreilly]

Summary

Moves ts-loader's development tooling from yarn to npm. This is an alternative to #1711 (pnpm), for comparison.

• Lockfile: yarn.lock is replaced by package-lock.json. It was generated by running npm install with the old yarn.lock present, so every resolved version matches yarn's.
• Peer deps: yarn never enforced peer ranges, and npm rejects babel-loader@7's webpack@2 || 3 peer. That is fixed with "overrides": { "babel-loader": { "webpack": "$webpack" } }. legacy-peer-deps was tried first, but it hoisted ajv-keywords@5 next to eslint's ajv@6 and broke every webpack build.
• Execution-test packs: all yarn.lock files are converted the same way, from empty node_modules. A script confirmed each package-lock.json resolves exactly the versions in its yarn.lock. The pnpm migration's pnpm import and npm's --package-lock-only both drifted.
  • run-tests.js installs packs with npm install --install-links --legacy-peer-deps. --install-links copies file: deps instead of symlinking them, which keeps the git-tracked copies in some packs' node_modules intact. The pnpm pack still installs with pnpm, as before.
• testLib: the comparison tests' npm link ./test/comparison-tests/testLib is kept. It is now native to the package manager and leaves the lockfile unchanged (checked).
• CI:
  • Workflows use cache: npm, npm ci, npm run …, and npm install -D … for the TS/webpack matrix. A dry run of both matrix installs resolves cleanly.
  • The benchmark job installs a base branch that predates package-lock.json with npx yarn@1, because npm can't install main as it stands (the babel-loader peer conflict).
• Other files: the Dockerfile, devcontainer, .vscode/settings.json, AGENTS.md, CONTRIBUTING.md and the test-pack READMEs are updated.

Not changed: the README's install instructions for ts-loader users (including Yarn PnP), REFERENCES.md, examples/, CHANGELOG.md, and the yarn hint in src/compilerSetup.ts.

Testing

Run locally on macOS with Node 24 and npm 11:

• npm run build and npm run lint pass.
• npm run comparison-tests: 89/89 suites pass.
• npm run execution-tests: 28/29 suites pass. 3.6.0_projectReferencesToBeBuilt also fails locally on main with yarn (karma temp-path issue on macOS), and it passes in CI.
• The benchmark job's base-branch steps were simulated against main, and the benchmark ran end to end.

🤖 Generated with Claude Code

[github-actions]

Benchmark (Ubuntu)

Scenario	transpileOnly	PR branch median (ms)	base branch median (ms)	Δ vs base branch
Cold build	false	961.3	925.2	+3.9%
Incremental rebuild (leaf touch)	false	62.0	60.8	+1.9%
Incremental rebuild (hub touch)	false	370.7	364.4	+1.8%
Cold build	true	566.4	576.5	-1.7%
Incremental rebuild (leaf touch)	true	31.1	32.0	-2.8%
Incremental rebuild (hub touch)	true	34.9	34.2	+1.8%

PR branch = /home/runner/work/ts-loader/ts-loader, base branch = /home/runner/work/ts-loader/ts-loader-main. 2 warmup + 10 measured iterations per scenario, median reported. Report-only - no threshold fails this check.

[github-actions]

Benchmark (macOS)

Scenario	transpileOnly	PR branch median (ms)	base branch median (ms)	Δ vs base branch
Cold build	false	1346.9	1321.0	+2.0%
Incremental rebuild (leaf touch)	false	87.0	81.9	+6.2%
Incremental rebuild (hub touch)	false	692.1	693.6	-0.2%
Cold build	true	871.8	865.2	+0.8%
Incremental rebuild (leaf touch)	true	59.5	56.6	+5.0%
Incremental rebuild (hub touch)	true	54.4	57.0	-4.6%

PR branch = /Users/runner/work/ts-loader/ts-loader, base branch = /Users/runner/work/ts-loader/ts-loader-main. 2 warmup + 10 measured iterations per scenario, median reported. Report-only - no threshold fails this check.

[github-actions]

Benchmark (Windows)

Scenario	transpileOnly	PR branch median (ms)	base branch median (ms)	Δ vs base branch
Cold build	false	747.5	754.9	-1.0%
Incremental rebuild (leaf touch)	false	53.5	53.3	+0.3%
Incremental rebuild (hub touch)	false	305.5	303.3	+0.7%
Cold build	true	518.0	526.9	-1.7%
Incremental rebuild (leaf touch)	true	29.7	32.5	-8.6%
Incremental rebuild (hub touch)	true	30.6	32.4	-5.6%

PR branch = C:\source\ts-loader, base branch = C:\source\ts-loader-main. 2 warmup + 10 measured iterations per scenario, median reported. Report-only - no threshold fails this check.