Struct FileStore 

pub struct FileStore {
    writer: Writer,
    compact_threshold: Option<u64>,
    bytes_since_compact: u64,
    frame_buf: Vec<u8>,
}

The runtime journal-file backend: an append-only writer plus compaction bookkeeping.

Fields

writer: Writer

compact_threshold: Option<u64>

bytes_since_compact: u64

Journal bytes written since the last compaction. Persists across reconnects with the store so short-lived connections do not reset it.

frame_buf: Vec<u8>

The single reused buffer the pending ops are framed into and then moved to the writer’s blocking pool, so a steady stream of flushes performs no per-flush allocation and no intermediate copy.

Implementations

impl FileStore

pub(crate) async fn persist( &mut self, client: &mut Client<JournaledSession>, ) -> Result<()>

Frames the session’s pending ops, appends them, and once the growth threshold is exceeded compacts the journal. Clears the in-memory buffer only on a successful append, so a failed write keeps the ops buffered and the returned error lets the caller stop acknowledging state that was never made durable.