Struct LogicalChangeApplier 

pub struct LogicalChangeApplier { /* private fields */ }

Shared logical change applier so replica replay and PITR converge on the same semantics. Stateful (PLAN.md Phase 11.5): tracks the last applied LSN + payload hash so duplicates / older LSNs / gaps / divergences are detected explicitly.

Implementations

impl LogicalChangeApplier

pub fn new(starting_lsn: u64) -> LogicalChangeApplier

Build a fresh applier. starting_lsn is the LSN already covered by the snapshot (or 0 for an empty replica). The next acceptable record is any positive LSN; from there the chain advances by 1.

pub fn last_applied_lsn(&self) -> u64

pub fn apply( &self, db: &RedDB, record: &ChangeRecord, mode: ApplyMode, ) -> Result<ApplyOutcome, LogicalApplyError>

Apply one logical change record. The state machine:

• first record after starting_lsn == 0 → apply, anchor.
• lsn == last + 1 → apply, advance.
• lsn == last && payload hash equal → idempotent skip.
• lsn == last && payload hash differs → Divergence (fail closed).
• lsn < last → older replay, skip with debug log.
• lsn > last + 1 → Gap (fail closed; caller marks unhealthy).

pub fn apply_record( db: &RedDB, record: &ChangeRecord, _mode: ApplyMode, ) -> Result<(), RedDBError>

Stateless apply — applies the record without monotonicity checks. Kept for callers that don’t yet thread the stateful applier through. New code should prefer LogicalChangeApplier::new() + apply().