Run & review this pull request in StackBlitz Codeflow.

Extracted the rollup/parseAst changes into #15639, so this one is easier to review and merge.

I'm not really oppose to this, but it's kinda unfortunate that we can't use the exported version now. Should we reach to Rollup if they could export it as a subpath? (though feels a bit redundant, i guess it would be nice if the utils are placed under rollup/utils perhaps)

I'm not really oppose to this, but it's kinda unfortunate that we can't use the exported version now. Should we reach to Rollup if they could export it as a subpath? (though feels a bit redundant, i guess it would be nice if the utils are placed under rollup/utils perhaps)

Do you mean that rollup should expose a small rollup/utils entry point with the version, and maybe later things like normalizePath? We can ping Lukas to see what he thinks later, but we could move forward with the PR and improve things later when we have a better story upstream.

Yeah I think we can figure out the rollup/utils thing later and we can do this for now. Sorry for not responding earlier, I'd prefer the esbuild changes done separately, or even not done at all 😬 I feel like we're using esbuild a lot more frequent than Rollup, and it's not really worth optimizing the esbuild import.

Yeah I think we can figure out the rollup/utils thing later and we can do this for now. Sorry for not responding earlier, I'd prefer the esbuild changes done separately, or even not done at all 😬 I feel like we're using esbuild a lot more frequent than Rollup, and it's not really worth optimizing the esbuild import.

I extracted the esbuild changes as a draft to

• perf: lazy load esbuild #15685

This PR only avoids statically loading rollup before the server starts now. I need to run proper benchmarks with the other extracted PRs so we can discuss them further 👍🏼