Struct ParallelWalLaneStager 

pub struct ParallelWalLaneStager<T> { /* private fields */ }

Production lane-local staging state for ordinary parallel WAL appends.

This keeps batch ownership, backlog accounting, and same-lane drain order inside fsqlite-wal instead of scattering the logic across pager-only callers. The caller still owns the final ordered durability residue.

Implementations

impl<T> ParallelWalLaneStager<T>

pub fn new(control: ParallelWalControlSurface) -> Self

pub fn control(&self) -> &ParallelWalControlSurface

pub fn next_batch_id(&self) -> u64

pub fn lane_count(&self) -> usize

pub fn current_lane_id(&self) -> u16

pub fn current_lane_backlog(&self, lane_id: u16) -> usize

pub fn record_batch(&self, batch: ParallelWalLaneBatch<T>) -> usize

pub fn take_batches_for_flush( &self, contexts: &[TransactionFrameBatchContext], ) -> Option<HashMap<u64, ParallelWalLaneBatch<T>>>

pub fn discard_batches_for_flush( &self, contexts: &[TransactionFrameBatchContext], ) -> usize

Discard prepared payloads for batches that already fell back to the raw ordered WAL path.

Lane-local preparation is an optimization, not the durability record. If a flusher cannot consume the prepared payloads for a group-commit epoch, those payloads become stale as soon as the same batches are appended via borrowed frame refs. Leaving them queued would make later epochs see an old batch id at the front of the lane and permanently fall back.

Trait Implementations

impl<T: Debug> Debug for ParallelWalLaneStager<T>

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.