Struct Enumerator 

pub struct Enumerator { /* private fields */ }

Stateful device enumerator: owns persistent channels and the per-device probe cache so the event-first watcher reuses immutable data across reconciliations. One-shot callers use the enumerate free function, which runs against a fresh (empty) cache.

Implementations

impl Enumerator

pub fn open_failures_last_tick(&self) -> bool

Whether the most recent enumerate pass failed to open at least one HID++ node. false before the first pass.

On macOS a run of passes with this set is the observable signature of a denied Input Monitoring grant or a stale permission session — paired with the grant state it separates “grant it” from “log out”, which the bare open error cannot (the denial is silent).

pub fn retry_needed_last_tick(&self) -> bool

Whether the last pass needs a bounded fast follow-up to advance a probe/channel/cache recovery policy. false before the first pass.

pub fn with_backend(backend: Arc<dyn HidBackend>) -> Self

An enumerator that walks backend — this host’s HID stack, a scripted device tree in tests, or another host’s.

pub fn with_registry(self, registry: ChannelRegistry) -> Self

Publish this enumerator’s already-open channels into registry after each settled inventory pass.

pub fn with_event_notifier(self, notifier: EventNotifier) -> Self

Subscribe this enumerator’s already-open channels to lifecycle events and coalesce them into reconciliation requests sent through notifier.

pub fn with_probe_cache(self, store: Arc<dyn ProbeCacheStore>) -> Self

Warm-start this enumerator’s immutable probe cache from store, and write it back there whenever its persistable content changes.

A modifier rather than a constructor: persistence is orthogonal to the channel registry and the backend, so an enumerator can carry all three.

pub async fn enumerate( &mut self, ) -> Result<Vec<DeviceInventory>, InventoryError>

One enumeration pass, reusing the cache from prior passes. Probes every HID candidate concurrently (so one asleep node that burns the whole PROBE_TIMEOUT can’t stall the others), reusing each device’s cached immutable data when it’s present and fresh.

A node the OS still lists but whose probe fails (receiver registers unanswered, probe timeout, open failure) is not reported as absent: its last completed inventory is replayed for a bounded grace and its channel is reopened, so a transient HID++ glitch can’t masquerade as “no devices” (#218) — see the node ledger.