Struct SyncEngine 

pub struct SyncEngine { /* private fields */ }

Decides when each registered target runs.

Implementations

impl SyncEngine

pub fn new( storage: Arc<dyn Storage>, clock: Arc<dyn Clock>, events: EventBus, ) -> Self

pub fn register( &self, target: SyncTarget, policy: SyncPolicy, source: Arc<dyn SyncSource>, )

Register a target. Registering the same target again replaces it.

Synchronous so that a module can register from ApplicationModule::register, which runs during startup and has no runtime to await on.

pub fn unregister(&self, target: &SyncTarget)

Stop tracking a target. Any run in flight is cancelled.

pub fn targets(&self) -> Vec<SyncTarget>

pub fn policy(&self, target: &SyncTarget) -> Option<SyncPolicy>

The policy a target was registered with.

pub async fn state(&self, target: &SyncTarget) -> Result<SyncState>

pub async fn due_at(&self, target: &SyncTarget) -> Result<OffsetDateTime>

When this target may next run.

Healthy targets follow their interval; failing ones follow the backoff; an offline machine gets the flat offline retry instead of an exponential one.

pub async fn run_due( &self, now: OffsetDateTime, ) -> Vec<(SyncTarget, Result<SyncOutcome>)>

Sync every target that is due at now.

Separate from the background loop so scheduling can be tested by moving a fake clock instead of by sleeping.

pub async fn sync_if_due( &self, target: &SyncTarget, ) -> Result<Option<SyncOutcome>>

Sync unless the target ran very recently.

This is the entry point for triggers that fire on their own — window focus, network coming back, a view being opened. Without the throttle, alt-tabbing twenty times means twenty syncs.

Returns Ok(None) when the run was skipped. A user pressing Refresh should go through SyncEngine::sync_now instead: they asked explicitly.

pub async fn sync_now(&self, target: &SyncTarget) -> Result<SyncOutcome>

Sync one target immediately, whatever the throttle says.

Single-flight: a second caller waits for the run in flight instead of starting a parallel one. Two concurrent syncs of the same target would race on the validators and could store an older result over a newer one.

pub async fn health(&self) -> Health

Health across all registered targets — the worst state wins.

pub async fn health_of_account( &self, connector: &ConnectorId, account: &AccountId, ) -> Health

Health of everything belonging to one account.

pub async fn run(&self, stop: CancellationToken)

Run the scheduler until stop is cancelled.

Returns a future rather than spawning a task: which executor runs it, and on which thread, is the host’s decision. A platform crate that called tokio::spawn itself would panic wherever no runtime is entered — which is exactly what a Tauri setup hook is.

A thin wrapper around SyncEngine::run_due; all the logic worth testing is in there, not in this loop.

impl SyncEngine

Convenience for the common “every interval” registration.

pub fn register_every( &self, target: SyncTarget, interval: Duration, source: Arc<dyn SyncSource>, )

Trait Implementations

impl Clone for SyncEngine

fn clone(&self) -> SyncEngine

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for SyncEngine

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

Formats the value using the given formatter.