Struct CompactionStore 

pub struct CompactionStore { /* private fields */ }

Implementations

impl CompactionStore

pub async fn open(db_path: &str) -> Result<Self>

Open or create the SQLite file at db_path. The directory must already exist (we don’t create it). Use :memory: for tests.

pub fn with_pool(pool: SqlitePool) -> Self

Wrap an externally-opened pool. Useful when sharing the same SQLite file as another store. Caller is responsible for migrate.

pub async fn migrate(&self) -> Result<()>

pub async fn insert(&self, row: &CompactionRow) -> Result<()>

Append a successful compaction. Errors only on storage problems.

pub async fn latest(&self, session_id: Uuid) -> Result<Option<CompactionRow>>

Most recent compaction for a session, if any.

pub async fn list_for_session( &self, session_id: Uuid, limit: u32, ) -> Result<Vec<CompactionRow>>

All compactions for a session, newest first, capped at limit.

pub async fn try_acquire_lock( &self, session_id: Uuid, holder: &str, ttl_seconds: u32, ) -> Result<bool>

Try to take the per-session compaction lock. Returns true when acquired, false when another holder is active. Stale locks (older than ttl_seconds) are evicted before the acquire attempt so a crashed compactor doesn’t deadlock the session forever.

holder is a free-form debug label ("pid:thread_id" works well) recorded on the row to make orphan diagnosis easier.

pub async fn release_lock(&self, session_id: Uuid) -> Result<()>

Release a previously-held lock. Idempotent.

pub async fn cleanup_stale_locks(&self, ttl_seconds: u32) -> Result<u64>

Sweep every lock older than ttl_seconds. Call from a background task or on boot to clean up after crashed processes. Returns the number of rows removed.