Struct SharedHead 

pub struct SharedHead<P> { /* private fields */ }

TTL-cached, single-flight HeadSource over one provider.

Implementations

impl<P: Provider> SharedHead<P>

pub fn new(provider: P, poll_interval: Duration) -> Self

Cache head reads for half the watcher poll interval.

Half, rather than a full interval, because the watchers are not phase-aligned: they bootstrap sequentially and their ticks drift across the interval, so it is the TTL — not the single-flight — that collapses the reads. A full-interval TTL would let a watcher act on a head an entire tick old, silently stretching the effective cadence of a documented knob toward 2×; half an interval bounds the added staleness below one tick for one extra RPC per interval. It also self-scales: an operator who drops event_poll_interval_ms to 250 ms for a local anvil gets a 125 ms TTL and a near-live head, with no second knob to tune.

pub fn with_ttl( provider: P, ttl: Duration, rpc_call_timeout: Option<Duration>, ) -> Self

Self::new with an explicit TTL and per-call timeout. Duration::ZERO disables caching (every call re-reads).

Trait Implementations

impl<P> Debug for SharedHead<P>

Hand-written rather than derived: P is a provider and carries no Debug bound, and the cached entry sits behind an async mutex this must not block on.

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

Formats the value using the given formatter.

impl<P: Provider> HeadSource for SharedHead<P>

fn head<'life0, 'async_trait>( &'life0 self, ) -> Pin<Box<dyn Future<Output = Result<u64>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait,

The current chain head, at or before the true head and never ahead of it. Err keeps its typed cause in its chain.