Rsbuild – A Better Vite?
rsbuild.dev
rsbuild.dev
I'm curious to see what Rsbuild brings on top of this already very competitive space of Rust-based builders (and I haven't even mentioned the fantastic non-Rust ones like ESBuild!)
I would personally bet on Vite at this time because it's the most complete package with the most momentum, and it's already hit some of the best speeds out there. Plus I am really rooting for a Rust-based package that compiles well as WASM because I am very inclined towards projects that do a lot of code analysis/compilation in the browser.
SWC and Turbopack aren't related in the sense that one is the "successor" of the other. They both do different things, and compliment each other. SWC is more lower-level, it's a compiler for JS (and other web tools) that converts your syntax into something any browser can understand. Turbopack is a bundler, it takes that compiled code and minifies/concatenates it together in various ways so it can be distributed to a browser efficiently. I believe Turbopack does in fact use SWC (if you're using Next that is) to do the "dirty" work of compiling TypeScript code into JavaScript quickly, but its main feature is the use of the Turbo engine to cache function calls at a very low level. From what I read, Turbopack's potential to make building JS apps incredibly efficient should be a really neat thing to work with in the future!
But is this constant turnover getting ridiculous in the web dev & JS ecosystem?
Yes. I simultaneously credit and blame facebook for starting it with yarn; after yarn, the trend of re-writing build tools seemed to become much more acceptable and trendy, especially since yarn (at the time) was multiple times faster than npm; you could argue it was worth it. But now we've reached a stage where the diversity of tools like npm, yarn and pnpm, now tools like ni (https://github.com/antfu-collective/ni) are emerging to manage the diversity.
I think this is getting out of hand, now there are n+1 of them.
I call this the "JavaScript treadmill", where everything that's more than two years old is garbage written by nincompoops, but now we all got it right and now everything is fantastic, and the only people who criticize it are grumpy crusty old backend boomers too lazy to learn anything new.
If you go back in time two years, people said the same thing. And four years, and six years, etc.
Who is giving that much money to an obscure build tool ? What is the monetization plan ? How are they ever trying to earn the money back ?
Does this happen in any way or form outside the US ?
The goal nowadays for these kind of tools is to get bought out by platforms such as Vercel/Fly/etc. or to become a platform of their own. Given that devs these days are afraid if not unable to deploy a React website on a VPS by themselves, it's become a self-sustaining lifecycle.
This tends to be a particularly US phenomenon. It's a house of cards waiting for the next shaky economic phase.
https://www.youtube.com/watch?v=E-w0R-leDMc - starts about halfway through.
It’s a build tool that swaps an underlying Webpack process with a Rust-powered process which aims to be API-compatible. With the Webpack ecosystem.
The Vite developers are working on the analogous Vite replacement, which is an API-compatible Rust rewrite of Rollup: https://rolldown.rs/
https://rsbuild.dev/community/releases/v1-0
And 30ms slower as vite at hmr is totally ruining my DX
Since I don't care if my build takes 1 second more (per their benchmark again), but I do care very much that the page refreshes faster when I do an edit, it's a hard sell.
Got a fair bit of attention 44 days ago (25 points) https://news.ycombinator.com/item?id=41502473
I live for the day bundlers are no longer needed.
The state of modern JS such that much of it does not work on the web by default is absurd.
All the work that's been spent building better and better bundlers would have been better spent fixing the blight that Node, CJS and the death of Bower caused to begin with.
How we ended up in a place where most JS is server-first and needs to be adapted for the web is beyond me.
I know it probably comes down to devs starting out with React and then thinking that that's how hard webdev is supposed to be. But the reality is that the vast majority of apps don't need to be SPAs with Reacts/Formiks/Flux-derivatives/Routers and would be faster to develop, offer better UX and have a single digit fraction of the boilerplate code to do the equivalent.
There is always the option of not building.
Maybe if your goal is to minimize build times, but my goal is minimizing iteration times. With a decent build tool, it is much faster to bundle up a bunch of JS files and serve them as a single file rather than letting the browser (or even node) figure that out. This is also my peeve about the vite dev server (that makes individual requests for each source file), which is slower on a full reload than plain esbuild.
The problem is using 99% of third party libraries without building has become nearly impossible.
To be fair, both are improvements over the existing tools, if not entirely compatible.
So I just went middle ground and ditched Babel for swc.