implement permission-set types

[jphastings]

Hey @tylersayshi I had a crack at implementing #82!

I kept everything in lib.ts to fit with your current layout, but it's starting to get a bit large, so let me know if you'd like things broken out.

I've also added a helper function for fields which have companion *:lang fields, so translations can be grouped with their default!

{
  title: lx.langString("The default", {
    "es": "El predeterminado",
    "ja": "デフォルト",
  }),
}

Becomes:

{
  "title": "The default",
  "title:lang": {
    "es": "El predeterminado",
    "ja": "デフォルト",
  }
}

Fixes #82

[tylersayshi]

overall looks good! thanks for taking this on 😺

and I'm not worried about lib.ts getting large - I don't expect it to grow much further and am in general in favor of having that single large file with the implementation.

[jphastings]

Thank you for the detailed review! I thoroughly agree — so much of the benefit of your package is great document in-IDE! I'll make the improvements and update shortly

[jphastings]

Hey @tylersayshi; I've addressed the points you made, though I'm struggling with maintaining both the key ordering in the snaps as you requested and ensuring the tests pass. If you have any advice I'm all ears!

[tylersayshi]

hey! sorry there was a setting blocking the CI from running (these should run automatically going forward, i changeed the setting). Pretty much everything looks good now, just some tests that need to get fixed.

One last thing i'd like this PR to add to is the benchmark test: https://github.com/tylersayshi/prototypey/blob/main/packages/prototypey/core/tests/infer.bench.ts

it's ok (and expected) if more instantiations are caused by this change, but if the number goes up significantly, i'll like to look into the type inference implementation closer.

[jphastings]

Sure thing @tylersayshi — the test failures are just the ordering issues, so I'll fix those now & work on the benchmarks.

[jphastings]

So I've added some benchmarks, and run the existing ones — I think this is only adding a small increase in instantiations. I'm not familiar with this analysis in Typescript, so I had to ask Claude to help out here, but here are it's results:

Benchmark	Baseline (main)	Branch	Δ
infer with simple object	685	718	+33 (+4.8%)
infer with complex nested structure	956	1011	+55 (+5.8%)
infer with circular reference	634	685	+51 (+8.0%)
infer with app.bsky.feed.defs	1237	1327	+90 (+7.3%)
fromJSON infer with simple object	438	463	+25 (+5.7%)
fromJSON infer with complex	499	524	+25 (+5.0%)
fromJSON infer with circular	411	436	+25 (+6.1%)
fromJSON infer with app.bsky.feed.defs	513	538	+25 (+4.9%)

What do you think?

[tylersayshi]

yep, the benchmarks look good 👍 thanks for building this!