Struct ChangeLog 

pub struct ChangeLog { /* private fields */ }

An in-memory change log with bounded retention.

Older events are evicted when the log exceeds capacity. The sequence counter still increments monotonically; clients pulling with an old cursor will see only what remains in memory (or should issue a full re-sync if their cursor falls off the back).

Implementations

impl ChangeLog

pub fn new() -> Self

Create a new change log with the default capacity of 10,000 events.

pub fn with_capacity(capacity: usize) -> Self

Create a new change log with a specific capacity.

pub fn has_seen_op_id(&self, op_id: &str) -> bool

Returns true if this op_id was already applied. Used by the push handler to short-circuit replays. Callers that observe true should NOT re-apply the change and should return success to the client.

pub fn remember_op_id(&self, op_id: &str)

Mark an op_id as processed. Safe to call multiple times. Evicts the oldest entry when the cache exceeds op_id_capacity.

pub fn append( &self, entity: &str, row_id: &str, kind: ChangeKind, data: Option<Value>, ) -> u64

Append a change event. Returns the assigned sequence number.

pub fn pull( &self, cursor: &SyncCursor, limit: usize, ) -> Result<PullResponse, PullError>

Pull changes since a cursor, up to a limit.

Returns Err(PullError::ResyncRequired) when the caller’s cursor has fallen off the back of the retention window — i.e. the cursor’s last_seq is lower than the oldest seq we still remember. Previously this case was silent: pull would return the surviving tail and advance the cursor, so the client converged to a state that skipped the evicted events entirely. That’s a permanent correctness bug; clients should instead do a full re-sync from entity list state.

pub fn len(&self) -> usize

Get the total number of events in the log.

pub fn is_empty(&self) -> bool