feat(cli, core): remove yarn lock-in, adds ni

[vaporwavie]

What's the purpose of this pull request?

Solves the yarn lock-in issue by predicting which package manager is preferred by the developer.

How it works?

It uses a ni wrapper created by us, that aims to predict which package manager is being used by the developer.

How to test it?

• Experiment it with npm
• Experiment it with pnpm
• Experiment it with yarn

References

ni: https://github.com/antfu-collective/ni

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Comments	Updated (UTC)
faststore-site	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	Jul 18, 2024 1:53pm

[codesandbox-ci]

This pull request is automatically built and testable in CodeSandbox.

To see build info of the built libraries, click here or the icon next to each commit SHA.

[gvc]

An overall question is: does this mean that people should use ni instead of yarn when developing for FastStore? (Think VTEX developers and not people creating stores)

[vaporwavie]

@gvc actually, ni is not a package manager itself, but rather a toolset to identify which package manager the developer is attempting to use. That said, if VTEX developers are using yarn, then ni will read the package manager as yarn. Same goes for bun, pnpm, npm. In this implementation, we're specifically integrating around the na command from it: https://github.com/antfu-collective/ni?tab=readme-ov-file#na---agent-alias

[eduardoformiga]

I tested it using the core project, and it looks like it worked as expected, although the sequence volta run yarn run before the commands looks uncommon to me

In the starter, I've got this error:

[vaporwavie]

@eduardoformiga re. yarn running through volta, which part of that behavior felt uncommon to you? Talking to @gvc, running volta for yarn and node are still commands that can be expected, even if we're removing the explicit lock-in from yarn. Was there something uneven that you saw during the command execution?

[eduardoformiga]

LGTM!

Now it worked as expected.

[eduardoformiga]

@eduardoformiga re. yarn running through volta, which part of that behavior felt uncommon to you? Talking to @gvc, running volta for yarn and node are still commands that can be expected, even if we're removing the explicit lock-in from yarn. Was there something uneven that you saw during the command execution?

Yes, that should be expected. I'm not used to seeing these two commands together before. But it's not a problem. Let's go ahead :)