Figure out a consistent globbing strategy

Describe the problem

In #2247, we switched to using micromatch (from globrex) to determine which files are included/excluded when creating packages. But we’re still using globrex via tiny-glob here: https://github.com/sveltejs/kit/blob/aad27f4ed3a2a57f68c6cb55076b9d3132bcbaac/packages/kit/src/core/create_manifest_data/index.js#L50-L58

As a result, our inclusion/exclusion logic is inconsistent, and we’re bundling two libraries with overlapping functionality.

Describe the proposed solution

Personally I’d love to not use micromatch. It’s a very big library for constructing regexes. If we do need to provide the extra flexibility, then should we take a leaf out of uvu’s book and just accept regexes in the first place?

export default {
  kit: {
    package: {
      exports: {
        exclude: /^internal/
      }
    }
  }
};

Alternatives considered

Instead of regexes, a function?

export default {
  kit: {
    package: {
      exports: file => !file.startsWith('internal')
    }
  }
};

Importance

nice to have

Additional Information

No response

[jiv-e]

Online docs could show some examples how to do basic exclusions and globbing with some battle tested library.

[Karlinator]

I can get behind using a function for exclusions, though it does make some simple things slightly more complicated. It also makes the method of matching very explicit for the developer, which is good. Globbing, for instance, is not always consistent between flavours/options.

@jiv-e GitHub pro tip: if you write the language name (e.g javascript) just after the opening triple backtick you get syntax highlighting!

This: ```javascript const something = ‘a string’; ``` becomes:

const something = 'a string';

makes it a little easier to read 😄

I haven’t tested your suggestion either, but it looks reasonable I think?