no-unresolved rule should not report for optionalDependencies

[sindresorhus]

Imported packages that are defined in optionalDependencies in package.json are optional and should not cause a violation.

[ljharb]

they're "optional" in that the overarching install won't fail if their installation fails; that's not the same thing as optional.

no-unresolved also doesn't skip devDependencies when NODE_ENV is production - its purpose is to warn when a require wouldn't work. If you're requiring an optional dep, just like you'd already have to wrap it in a try/catch, you'd also need to add an eslint override.

In the linked use case, certainly the "windows.js" entry point might be optionally required, but unless the dependency is present, it'd fail at runtime.

The only thing I can think of that might possibly make sense here is an explicit option to skip optionalDependencies, or, an explicit "exceptions" array - but in general I think that conditional requires and optional deps run counter to the purpose of the rule.

[sindresorhus]

The point is that if I put a dependency in optionalDependencies, I'm explicitly indicating I want to handle it myself. I think this plugin should respect that, at least through an option.

[abcdapply]

I have a Jest test file that uses the redux-mock-store package, which gets installed with --save-dev. I get a false positive for import/no-unresolved and import/extensions