Explore moving to Vite (or Next.js?)

[vacekj]

Hello team,
I've spent some time to try out if a move from CRA to Vite would make sense for us. My main reasons for moving on from CRA were the following:

• CRA development has slowed down considerably, it took 14 months to go from 4.0 to 5.0. There hasn't been a new release since April.
• CRA is slow. It's not taking advantage of the latest developments in the ecosystem and tooling, namely ES Modules, esbuild or swc. It still uses Babel and Terser for transpilation and minification, which are orders of magnitude slower.
• CRA is not extensible by default. We are already using Craco though, so this is not much of an issue.

I've set up a branch with our repo converted to work with Vite here. The latest commit also tries to add support for Vitest, which was unfortunately much harder than I thought. I have only briefly tested the app itself, but the build and typecheck passes.

The build time went from 36.79s to 18.92s (including tsc, which is very slow), 12.73s without tsc. So an almost 2x improvement in build time.

However, there is another, lower-hanging fruit that we can pick - keeping craco, but swapping out Babel and Terser for esbuild. Branch here. This requires changing only the craco config and installing the package.

With esbuild, the build time is 14.78s, including type-checking, almost a 2.5x improvement in build time, with no code changes necessary. Pretty nice 😎

For now, I would just merge in the esbuild plugin, and scratch the Vite plan.

In the future, I would explore using Next.js, since they are the most well-funded and have the best feature set, and their shipping speed is extraordinary. They are making big advances on the performance front too by moving their compilation pipeline to Rust. But that is a discussion for after we launch the v1 I think, because it would be a larger architectural change.

[vacekj]

Closing as stale. we are now using craco esbuild plugin on all frontends, and migration to nextjs is underway for explorer.