API used in the plugins developed by BenCodez, can be used in any project.
<repository>
<id>BenCodez Repo</id>
<url>https://nexus.bencodez.com/repository/maven-public/</url>
</repository>
<dependency>
<groupId>com.bencodez</groupId>
<artifactId>advancedcore</artifactId>
<version>LATEST</version>
<scope>provided</scope>
</dependency>
repositories {
maven { url "https://nexus.bencodez.com/repository/maven-public/" }
}
dependencies {
compile "com.bencodez:advancedcore:LATEST"
}
Versions:
LATEST - latest stable release
Check out all tags on the releases tab.
Calendar checks requested by reload run on the existing background calendar timer, so asynchronous date events are not called inline from a Bukkit or Folia owner. Repeated requests coalesce while one check is queued or running. Shutdown waits for admitted checks; queued callbacks skip a retired checker or replaced timer. Normal periodic checks remain unchanged. This does not move the remaining reload file reads off the caller, or make reload completion await all date-event listeners.
When an asynchronous onBeforeStorageShutdown() stage settles, storage retirement
is queued on the existing user storage worker. The completing callback can finish
its admitted storage work before the shared cache flush and native provider close.
Already completed default hooks retain their existing retirement-stage behavior.
A rejected continuation reports failure and leaves the shared runtime attached;
it does not report a successful flush. Failed or cancelled preparation still
retires accepted writes, with the preparation failure retained. No new executor,
configuration, or storage format is introduced.
Shutdown captures user-storage admission and its manager before waiting for the calendar or consumer hook. Deferred unload can disable user loading or replace the public manager reference without skipping that admitted manager's flush and cleanup. The native provider is still resolved after shared retirement so a provider replacement completed by admitted work closes its current owner.
If consumer preparation exceeds the default five-second shutdown watchdog, the watchdog reports it and retains the existing daemon storage worker and native provider. Preparation is acknowledged only after its callback has submitted the final retirement task. A future becoming complete alone is insufficient because its callbacks can still be running. After that acknowledgement, a stalled storage retirement still receives the bounded watchdog and worker grace; terminal native cleanup waits for the actual storage worker to terminate.
The first watchdog fences an outstanding calendar transition and immediately interrupts its held calendar executor, even while consumer preparation remains pending. Interruptible calendar listeners can therefore retire without waiting for that hook. Storage admission and its native provider remain available for the eventual checkpoint; interrupting a listener does not prove its work persisted.
A consumer hook that never settles retains its storage owner until JVM exit. This does not block the server thread or keep the JVM alive, but an abrupt exit cannot guarantee the pending checkpoint was persisted. Consumers must complete, fail, or cancel their preparation stage. No new executor or polling loop is added.