peersync

func ApplyEvents ¶

func ApplyEvents(
	ctx context.Context,
	j journal.Journal,
	views []journalview.View,
	byID EventByID,
	events []journal.Event,
) (applied int, err error)

ApplyEvents appends peer events into the local journal and views, skipping any event whose id is already present (idempotent under pull replay). Returns how many events were newly applied.

func RunPullLoop ¶

func RunPullLoop(
	ctx context.Context,
	logger *zap.Logger,
	selfID uuid.UUID,
	members MemberSource,
	syncer PeerSync,
	j journal.Journal,
	views []journalview.View,
	byID EventByID,
	cursorStore CursorStore,
)

RunPullLoop is the peer journal reconciliation loop.

It does not push our events. Each tick it discovers who is alive via memberlist and pulls their journals over the Noise transport, up to maxPagesPerTick pages per peer while each page comes back full:

• On meet: peer is newly seen or has become alive again → Sync once now.
• Periodic: every defaultPullInterval, Sync all still-alive peers (skipped if already synced this tick for a meet).

After each successful Sync, events are applied idempotently (skip known ids) into the local journal and views. The cursor advances only if apply succeeds, in memory and in the store together.

Cursors persist in bbolt through the store, so a restart resumes mid-history instead of re-pulling from event zero. A store load failure starts from empty cursors instead; idempotent apply discards the overlap. A nil store keeps RAM-only behavior and exists for unit tests.

Individual Sync/apply failures are soft-fail (log and continue). The loop blocks until ctx is cancelled.