replace tshy in module-replacements

[kravetsone]

Why u use it as package builder?

https://www.npmjs.com/package/tshy?activeTab=dependencies

It uses the packages that you say should be replaced...

I suggest use pkgroll instead

[43081j]

it does a fairly good job. the usual two choices for dual packages are tsup and tshy from what i've seen

tsup bundles the sources twice (which looks like what pkgroll does too), and tshy just builds via tsc twice

are you concerned about the dependency tree depth?

[AbhiPrasad]

tshy is great, I don't think it's a huge deal that the packages in the module-replacements list are being used because rimraf and similar is most just there to deal with glob. Here is an example: https://github.com/isaacs/sync-content/blob/acdce90c55eecc603945c935bffbbd3d377f1e2e/src/index.ts#L219-L220.

tshy also doesn't use rollup, which does help keep it light.

pkgroll uses an outdated version of glob, which brings in the deprecated inflight, so feels like a worse option atm. It also brings in the whole rollup ecosystem, which is heavy in it's own right.

bunchee is another similar all-in-one dual emitting solution that I really like, but it has an even bigger dep tree than pkgroll (and also is rollup-based)

Then there is unbuild which I like so much I even did a conference talk about using it, but it has the biggest dep tree of all (why load cssnano for just bundling a simple isomorphic js lib 😭). It's the most extensible of all the options (with escape hatches to run custom rollup plugins), so it's great for more complicated monorepos like what I deal with at my day job.

[43081j]

makes sense

i think lets keep this on tshy for now with the ultimate aim of just moving to esm-only (can't really until eslint can load esm plugins)

unbuild is more than we need here but @pi0 and @danielroe seem pretty interested in the e18e stuff, so im sure they'd be open to suggestions for making it lighter 👀

[pi0]

Thanks for feedback and kind words dear @AbhiPrasad @43081j. Let's track it upstream: unjs/unbuild#405