Reinventing Rollup
summary
Rollup introduces concrete wins: the code‑splitting algorithm drops from 2 hours to a few seconds on a huge project; dynamic imports get tree‑shaking; a no‑side‑effects annotation marks functions once instead of per call; experimental chunk‑count reduction, side‑effect logging and a manual pure‑function override help shrink and reorganize output. Plugin API adds pre‑built chunks, emit‑only‑if‑referenced assets, a resolvedBy resolver trace and a richer logging API. Parser work swaps Acorn for SWC via Rust, hitting 51 ms native parsing and a smaller binary buffer for faster caching and future parallelism.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 1:05Code splitting improvements
- 2:10Tree shaking features
- 3:58Experimental side effects tools
- 5:23Plugin author features
- 7:54New logging API
- 9:21Performance bottlenecks
- 11:10Replacing the parser
- 13:45Binary syntax tree conversion
- 15:34Performance gains and caching
- 16:37Next steps in Rust
- 18:04Cost and availability
- 19:29Roll down project comparison
- 21:13Future features dream
related talks
45mBundler Bonanza
29mModern JavaScript Tooling: Exploring Rolldown, Oxc, and Beyond
39mBundlers: a deep dive into modern JavaScript build tools
22mOwning your Build-step – Owning your Code
25mRolldown: How Vite bundles at the speed of Rust
34mBuild-time Optimizations in JavaScript Applications
18mThe hashing dilemma, Rollup 3, and our future with Vite
35mAshley Williams - SeattleJS Conference 2017