Tech Stack: Roadmap

[carlosthe19916]

For the upcoming releases I thought we could evaluate some things that might be beneficial in the long term.

• https://vite.dev/ Evaluate if it is worth adopting it and what benefits we can get from it vs sticking with our current webpack configuration
• React Router TanStack: I anticipate more complex requirements in regards of handling URLs will come. Would we gain benefits adopting it. Things like type safety, etc.
• Tanstack Table. Could we gain something replacing our current table system with this? The idea should be to delegate as much as we can to this kind of libraries and then focus only on delivering features without being distracted by tools/utilities like this one.
• Eslint. We are not currently enforcing Eslint rules. We could start enforcing them through our CI pipelines and keep the code safe.

[gildub]

We definitely need Eslint so I created an issue to be followed by a PR to address it.

Regarding moving on a new builder we could definitely consider vite and also compare with Rspack if you want Webpack compatibility with Rust speed.
But if we're taking a high view on the project I would suggest to think about moving from nodejs to bunjs because the bundler is built in and in typescript.

[carlosthe19916]

An update of Vite vs Rspack: I went through the experience of migrating to Vite and here I have a fully migrated version using Vite #355 . That is for us understanding what migrating involves. I will find some free time to do the same but using Rspack so we can make a fair comparison.

So far Vite:

• Great documentation not just on their website but all over the internet people have resources written
• Easy to migrate
• Great ecosystem, all the plugins I could think of were there.

[carlosthe19916]

Another Update: Here is another PR for Rspack/Rsbuild #365

• Rspack vs Rsbuild.- Both the same but Rsbuild comes with default setup.
• It has limited documentation outside their website
• I didn't measure it but it feel faster than Vite
• Configuration very similar to Vite.

[gildub]

Both Vite and Rsbuild are making the whole build process lighter and about the same from that perspective.
Rsbuild seems even a bit more efficient than Vite.

With no experience with both it seems difficult to cut between them as there're modern and fast.

Meanwhile after doing a deeper comparison,

Rsbuild seems more suited for Trustify-UI because of its Webpack excellent compatibility including with plugins which is perfect for us as we're migrating a "complex" Webpack project.

Rsbuild performance seems to exceed Vite and a maximum build performance is surely an important element for this project.

The two other key elements, Rsbuild has a strong module federation support and offers versatility. Effectively Rsbuild is designed to be a universal build tool, adaptable to various project types.

Conversely Rsbuild is a bit younger than Vite so it's community seems smaller.

Another important part are regarding the linters.

Vite is flexible and therefore the classic ESlint/Prettier can be still used as before (or almost ;) so that's not disruptive.
On the other side, Rsbuild comes with Biome which is very fast (implemented in Rust too) and brings a modern approach which simplifying/consolidate the code validation. This seems to be a real evolution.

Together Rsbuild and biome provides less configuration overhead.

So on the paper and from the little I have seen about them, I'm strongly in favor of Rsbuild.

But what do you think and what's your experience feedback after having dealt with both ?

[carlosthe19916]

I agree with all the points you mentioned:

• Vite => big ecosystem and community . RsBuild => is growing but not at the Vite's level yet
• Rsbuild is the winner in performance.
• Rsbuild comes with Biome but we can still configure to use Eslint + Prettier if we want to opt-out of Biome. I like the fact that they don't enforce it but let the decision to developers.

I was gladly surprised by the similarities Rsbuild has compared to Vite. It is like Rsbuild was heavily inspired by Vite when Rsbuild was designed. So if anyone has experience with Vite, then it should be straight forward to deal with Rsbuild.

Like you, I am also inclined to go for Rsbuild rather than Vite given its performance advantages.

I will polish the Rsbuild and remove the draft status.

The only question I do have is "when is the best time" for merging the Rsbuild migration. As we are approaching v2.0 I am kinda a conservative guy who doesn't like to add risk to releases. But I guess there is always gonna be risk all the time :)

[gildub]

@carlosthe19916,

Looking forward to review the final version with Rsbuild then !

About the "when" question, I suppose it depends when you can push the final version.
If you can do that by early next week then we still have time to validate it.
Let me know.

[carlosthe19916]

Closing this one:

• We moved to use Rsbuild
• Stayed with react-router-dom
• Avoid Tanstack Table and prefer the custom internal table handling