Struct BatchWorker 

pub struct BatchWorker { /* private fields */ }

Batch worker state.

Manages batch assembly for a single shard.

Implementations

impl BatchWorker

pub fn new( shard_id: u16, config: BatchConfig, next_sequence_published: Arc<AtomicU64>, producer_nonce: u64, ) -> Self

Create a new batch worker.

next_sequence_published is the bus-owned mirror of next_sequence. Pass Arc::new(AtomicU64::new(0)) if the caller doesn’t need to observe the post-exit sequence; production paths share it with bus::remove_shard_internal.

producer_nonce is stamped on every produced Batch for cross-process dedup. The bus passes its loaded nonce in; tests can use any u64 (typically 0 or the per-process default).

pub fn add_events(&mut self, events: Vec<InternalEvent>) -> Option<Batch>

Add events to the current batch.

Returns a completed batch if thresholds are met, or None if more events are needed.

Empty-input side effect

Passing an empty events vec is not a no-op. The BatchWorker’s recv-timeout arm calls add_events(vec![]) specifically to drive a check_timeout round, which may flush the in-memory current_batch if max_delay has elapsed since the last event arrived. Callers who want “true no-op on empty input” must check events.is_empty() themselves before calling.

Pre-fix this side effect was not documented and surprised callers expecting add_events([]) to be inert. The fix is documentation only — the BatchWorker’s timeout flush relies on this behavior, so removing the side effect would break the timeout-flush mechanism in bus.rs.

pub fn check_timeout(&mut self) -> Option<Batch>

Check if the batch should be flushed due to timeout.

Pre-fix perf #38 in docs/performance/net-perf-analysis.md this was private and the bus’s timeout branch called add_events(vec![]) as the indirection — an empty Vec allocation per timeout tick purely as a sentinel. The method is now pub so callers can express the intent directly without the alloc and without the documented “empty add_events has a side effect” footgun.

pub fn flush(&mut self) -> Batch

Force flush the current batch, even if thresholds aren’t met.

pub fn has_pending(&self) -> bool

Check if there are pending events.

pub fn pending_count(&self) -> usize

Get the number of pending events.

pub fn target_batch_size(&self) -> usize

Get the current target batch size.

pub fn time_until_timeout(&self) -> Option<Duration>

Get time until the current batch times out.