Get rid of Guava

[JFronny]

As suggested on Discord, this PR gets rid of the dependency on Guava.
Mostly, this meant switching code to using Jetbrains annotations for nullability, using the constructors for various collections directly, and replacing Preconditions.checkNotNull with Objects.requireNonNull.
Some things required special changes that may be a bit more controversial:

• the single usage of Cache in CachingClassProvider was replaced with an ad-hoc cache implementation
• Multimap usages were replaced with Map<K, Collection<V>> and computeIfAbsent
• The single usage of Streams.zip was replaced with a custom Iterator implementation
• I18n was modified to use an index generated by Gradle instead of doing its own reflection
• EntryTree was modified to extend Collection instead of just Iterable to allow using cleaner construction of copies (Collections constructors do not accept simple iterables)

These changes do have potential to break stuff, so I have quickly checked that things seem to work. If someone that has more experience with this codebase could take a look at them, that would be appreciated.

[Earthcomputer]

Overall looks good. Make sure you run checkstyle, there are some places where if statements don't have braces.

Edit: not sure how checkstyle passed, but anyway if (condition) throw new IllegalArgumentException(); should have braces.

[Earthcomputer]

This check should also be moved to something more careful like the synchronized block above, otherwise multiple threads could pass this check and remove too many entries from the cache. Although I'm questioning whether we even need to remove from the cache if it gets to big, maybe we should only rely on the expire time.

[JFronny]

I was of the opinion that removing one more element in case of a race probably isn't that bad and is probably better than the overhead of an additional synchronization. With that said, I have now added synchronization. One minute is quite a long time, so I'm not sure if removing the check is a good idea.

[Earthcomputer]

Nit: this class can be a record

[JFronny]

Unfortunately, it can't, as addTime needs to be updated if the cache value is accessed and that isn't possible with records (I could use an Atomic though)

[JFronny]

not sure how checkstyle passed, but anyway

the specific case of checking an argument in an if and throwing an IllegalArgumentException already exists without braces in Command since 2023, so if that failed checkstyle, the current master would probably fail

[Juuxel]

The checkstyle config allows single-line ifs without braces.

Enigma/checkstyle.xml

Lines 82 to 86 in 9abdbd5

	<!-- single line statements on one line, -->
	<module name="NeedBraces">
	<property name="tokens" value="LITERAL_IF,LITERAL_FOR,LITERAL_WHILE"/>
	<property name="allowSingleLineStatement" value="true"/>
	</module>

[Earthcomputer]

I personally think that should be changed in the checkstyle in a future PR, at least I don't think we should be adding new code with single line if statements.

[modmuss50]

I personally think that should be changed in the checkstyle in a future PR, at least I don't think we should be adding new code with single line if statements.

I agree, I wasnt even aware that was allowed, I thought it might have just been an enigma thing, but no its also in FAPI.

[modmuss50]

Awesome, many thanks.