Language Refactor 3

[Mnemotechnician]

Description

This significantly improves the quality of the language system by fixing the mistakes I've made almost a year ago while developing it.

Mainly, this throws away the old half-broken way of networking in favor of the component state system provided by RT. Language speaker comp is now shared with SendOnlyToOwner = true, and its state is handled manually.

In addition to that, this brings the following changes:

• UniversalLanguageSpeaker and LanguageKnowledge are now server-side
• DetermineLanguagesEvent and LanguagesUpdateEvent are now shared (so that future systems can be built in shared, if needed)
• Everything now uses the ProtoId type instead of raw strings (god, I hated those so much)
• The server-side language system now accepts Entity<T?> arguments instead of EntityUid + T
• UniversalLanguageSpeaker is now based on DetermineEntityLanguagesEvent and gets an Enabled field, which allows to turn it off. This may have some use in the future.
• Some minor cleanup

Changelog

No cl

[VMSolidus]

Suggested change

	public List<ProtoId<LanguagePrototype>> SpokenLanguages = [];
	public List<ProtoId<LanguagePrototype>>? SpokenLanguages = default!;

We really should be doing these as Nullable lists and not a default empty list. Anywhere you logically check "If empty return", you can instead check "if null return". This is actually overwhelmingly preferred now according to Microsoft's own documentation, ever since Dotnet 8 added the ability for IDE's to check for null references before compiling. The given reason is that it offers a substantial performance improvement, while still using the same general logic structure.

[Mnemotechnician]

I don't really see the benefit, but I see how it would be highly inconvenient for every caller to check those fields for null, when in reality they would never be null. The performance difference should be very negligible here, although if it concerns you a lot, I simply can change the initializer to default!, without changing the field type.

[Mnemotechnician]

Everything works as far as I could test, so marking as ready for review.

[VMSolidus]

And that's why I merge master before testing PRs now. Excellent. This PR doesn't work. :)

[Mnemotechnician]

merge is unrelated, I broke the linter somehow.

[Mnemotechnician]

The linter was unhappy I had an empty string as the default value for LSC.CurrentLanguage. Well, I revoked its right to check that field.

[VMSolidus]

Can't approve this until you add back in the functionality for removing UniversalLanguageSpeakerComponent also removing the Universal Language. Here I have a Cataloguer that has been mindbroken(which removes the UniversalLanguageSpeakerComponent), but even after the component is removed, the Cataloguer retained their ability to speak Universal.

[Mnemotechnician]

Can't approve this until you add back in the functionality for removing UniversalLanguageSpeakerComponent also removing the Universal Language. Here I have a Cataloguer that has been mindbroken(which removes the UniversalLanguageSpeakerComponent), but even after the component is removed, the Cataloguer retained their ability to speak Universal.

Mmmm, this just means the xenoglossy system is broken as it does not call UpdateEntityLanguages when adding/removing language-related comps.

I'll make it get called automatically on the comp removal event... later, when I get time.