craco VS parcel

I had some 'fun' figuring out how to deal with not going through create-react-app without doing a full eject, got something barely working ... and discovered https://craco.js.org/ already existed and did precisely what I'd part-implemented except better.

On the upside, by that point I knew the CRA codebase well enough to predict what it would do even in edge cases, and CRACO's implementation was immediately comprehensible, and none of my colleagues had to try and understand my half-arsed NIH version.

(avoiding being in any of this situation in the first place would likely have been preferable, but given where things were when I landed on the project in question that would've required a TARDIS)

If you initiated your React application using create-react-app, leverage @craco/craco to override your webpack configuration.

Ejecting a React app is a bad idea because our application will lose all the React configurations and will not benefit from the CRA updates. Some solutions for ejecting our application include using packages like react-app-rewired or rewire. You can also use CRACO to eject your React application, but it needs you to install additional plugins.

So, instead of entirely managing these configuration files, teams took to utilizing tools such as Craco to override configurations. These tools also come with their limitations: they were not updated as quickly as CRA, so there was always a lag in implementing new features, and they added an extra layer of complexity to existing tools through overrides and additional tools.

I am not much of a fan of CRA myself but I am very much glad that https://craco.js.org exists - so far it's handled my needs for tweaking CRA behaviour in situations where "eject" didn't seem like a good route to take.

Mostly tbh to stop the freaking thing spawning inotify watchers for the entire contents of node_modules - I don't mind having to do a manual restart when I've changed dependencies and I definitely -do- mind having it eat a shedload of my user's inotify kernel allocation. (I know you can up the allocation, that's not the point, why are you on my lawn? :)

As I mentioned, in our case, the perfect tool for this job is CRACO. It will let us simply overwrite CRA’s configuration without ejecting.

Create react app + Craco

Or I could use Craco with Craco-less. Craco 6+ doesn't work with 5+ versions of react-scripts. I know I could use yarn instead of NPM which doesn't stop the installation of craco, but it can't be the solution. We can't scrap the project and restart. Further, Craco 7-alpha installs but then craco-less doesn't.

If you want to customize the webpack configuration, you either need to eject, or to work against the package (with yarn patch, forking react-scripts, or using CRACO which is the easiest). But none of them are officially maintained by the CRA team.

It runs using Parcel, very simple and easy to setup. The app has 3 files:

Homepage: https://parceljs.org/

In the Changelog Podcast episode referenced above, Dan Abramov alluded to Parcel working on RSC support as well. I couldn’t find much to back up that claim aside from a GitHub issue discussing directives and a social media post by Devon Govett (creator of Parcel), so I can’t say for sure if Parcel is currently a viable option for developing with RSCs.

Parcel 2 emphasizes a zero-configuration approach to bundling web applications. It's a powerful tool that offers a hassle-free developer experience, focusing on simplicity and speed.

Once you build a simple Vite backend integration, try not to complicate Vite's configuration unless you absolutely must. Vite has become one of the most popular bundlers in the frontend space, but it wasn't the first and it certainly won't be the last. In my 7 years of building for the web, I've used Grunt, Gulp, Webpack, esbuild, and Parcel. Snowpack and Rome came-and-went before I ever had a chance to try them. Bun is vying for the spot of The New Hotness in bundling, Rome has been forked into Biome, and Vercel is building a Rust-based Webpack alternative.

Parcel

I’ve tried something similar on the frontend side: I decided to build a UI for Ollama.ai using only HTML, CSS, and JS (Single-Page Application). The goal is to learn something new and have zero runtime dependencies on other projects and NPM modules. Only Node and Parcel.js (https://parceljs.org/) are needed during development for serving files, bundling, etc. The only runtime dependency is a modern browser.

Here's what I have found so far:

- JavaScript (vanilla) is a viable alternative to React.js

Besides Webpack, there are many other popular web bundlers available, such as Parcel, Esbuild, Rollup, and more. They all have their own unique features and strengths, and you should make your decision based on the needs and requirements of your specific project. Please refer to their official websites for details.

In the Node.js ecosystem, bundling is typically handled by third-party tools rather than Node.js itself. Some of the most popular bundlers in the Node.js world include Webpack, Rollup, and Parcel, offering features like code splitting, tree shaking, and hot module replacement.

There are projects attempting to do more things. I've really enjoyed Parcel (https://parceljs.org). But it won't handle things like linting or unit testing, which you may or may not want. Vite is also pretty popular (https://vitejs.dev/), and it has a test runner.

Thing is, most of the problems described in the post aren't related to low-JS front-end libraries like HTMX or alpine. You can write React without a linter, bundler, build tool, unit testing, or linting. But with any of these projects at scale, you start wanting more:

- If you want to write unit tests in JS, you need to choose a test runner (probably Jest or Vitest -- until the built-in node testing module becomes more common).

- If you want linting, you need a linter (probably Eslint). If you want type safety, you need a type checker (probably Typescript).

- If you want to create smaller JS files to ship to production and to automatically handle assets, you need a bundler.

- If you want to use new language features while supporting old browsers, you need polyfills.

- If you want to use all these things together, you need something to bring it together (like Webpack).

So it really depends what you need! You may not need any. But as you can imagine, in many professional projects with multiple developers it's very nice to have unit tests, linting, and type checking :) (And you start caring about end-user performance a lot more, in which case optimizing the shipped bundle is important.)

Take all that, and then compare to a language like Rust, which has most of the "ecosystem stuff" built-in. In Rust, you get the test runner, the linter, dependency manager, type checker, and documentation tool all included. Easy! Thankfully, Rust doesn't have to care about whether users support modern language features (because it compiles down to lower code ahead of time), or whether the binary shipped to the client is optimally organized for downloading immediately over the internet.

It's a problem in JS because A) you have to care about more problems than many other languages since JS needs to load instantly over the wire in a web browser, and B) there is a huge amount of choice and not a lot of standardization in web tools. (And what standardization there is (Node, npm), there are still competitors trying to even further reduce the pain points.)

I think that in ten more years, we'll be in a better place, because there is push back (like this post!) against these problems, which will encourage more tools trying to solve the explosion of tools. Which seems counterintuitive, but these tools were created to solve very real problems. So I see it as a pendulum which has swung too far, but will likely swing back to a more balanced place. And you see that with tools like Vite gaining popularity.