✅ Deploy Preview for rspack canceled.

Built without sensitive environment variables

!canary

canary release moved to https://github.com/web-infra-dev/rspack/actions/workflows/release-canary.yml.

!bench

📝 Benchmark detail: Open

Threshold exceeded: ["threejs_development-mode_10x + exec","threejs_development-mode_10x_hmr + exec","threejs_production-mode_10x + exec"]

Generally looks good to me, the benchmark shows in an actual case (arco-pro) this is acceptable, but performance has regressed significantly when there are lots of javascript module (threejs 10x), can this be further improved?

@ahabhgk
The overheads could be reduced from two parts:

1. In development and production, it needs to parse all modules at the first build. I think the most time cost on AST parse from the scratch, I took a look in getting AST from module but didn't find. Is there a way to do that?
2. In development, we could use a parsed result cache map and reuse only re-parse the module only when modified.
   Any other suggestion?

There isn't a way to get the AST from the module, we don't store the AST on the module, we analyze dependencies from AST (for javascript) then drop it, especially you need to parse the generated code of the module (source has been modified from the original source, so does AST)

For caching, something like https://github.com/webpack/webpack/blob/34e2561addb0f65a7a6fb0ce7ae1aea4cd1d599f/lib/javascript/JavascriptModulesPlugin.js#L592 but works for caching the inline module result looks good to me

!bench

📝 Benchmark detail: Open

Threshold exceeded: ["threejs_development-mode_10x + exec","threejs_development-mode_10x_hmr + exec","threejs_production-mode_10x + exec"]

@ahabhgk
The benchmark got better, especially in HRM case. I tested it locally, the upper is with cache and below without. Seen that following HMR speed got faster as cache hits.

Something wrong in our CI, @SyMind is looking at it

Patched a commit lost in force-pushing that fix the wrong usage of ReplaceSource, test cases're updated alongside.

@SyMind Still failing with a retry just now, is this CI issue still being resolved? 🥲

@fi3ework I am trying to update the branch to get the latest code. Please wait for the CI to run again.

@ahabhgk The review was dismissed, PTAL again.