Struct RepeatingQuery 

pub struct RepeatingQuery { /* private fields */ }

A declared querier carrying the same RFC 05 §2.1 discipline as fleet_get (target All, consolidation None, attribution by reply key), for fetches that re-ask the same key expression — watch loops, the schema cache’s re-asks, registry sweeps, doctor. Declaring once lets the network keep routing state warm instead of rebuilding it per GET (report §12’s zenoh-1.9 adoption row).

When to use which:

• recurring, same keyexpr → declare a RepeatingQuery and fetch many times (parameters and payload ride per get, never in the declared keyexpr — a ?params suffix in key is a bug here);
• genuinely one-shot, or an ad-hoc key → fleet_get.

Liveliness sweeps (crate::bus::roster::roster()) are a different API (session.liveliness().get()) with no querier equivalent and stay undeclared.

Implementations

impl RepeatingQuery

pub fn key(&self) -> &str

The declared key expression.

pub async fn fetch(&self) -> Result<Vec<FleetAnswer>>

One fetch on the declared keyexpr, every reply attributed by its own key — fleet_get’s contract, minus the per-call declaration.

pub async fn fetch_with( &self, params: &str, payload: Option<Vec<u8>>, ) -> Result<Vec<FleetAnswer>>

As fetch, with selector parameters and/or a request payload riding this one get.

pub fn max_replies(self, max: usize) -> Self

Keep at most max replies per fetch (#339). Zero is clamped to one.

pub fn reply_bound(&self) -> usize

The reply bound in force.

pub fn elided(&self) -> u64

Replies this querier’s bound refused, across every fetch (RFC 13 §3 O6). See GetOpts::elided for why the count lives with the bound.

pub fn reset_elided(&self)

Forget what earlier fetches through this querier cost — for a caller that re-runs a sweep and reports per sweep rather than per querier (GetOpts::reset_elided is the same call on the one-shot path).

Without it a per-sweep figure has to be read as a before/after subtraction, which is not safe when two sweeps overlap on one declared querier.

pub async fn fetch_timed(&self) -> Result<Vec<(FleetAnswer, Duration)>>

As fetch, stamping each reply with how long after the GET it arrived (issue #52).

This exists because a fan-out call’s call duration is the time until the slowest answer, so attributing it to every origin would report a fast responder’s latency as the fleet’s worst. Timing each reply where it is drained is the only place the distinction is available — and it keeps the RFC 05 §2.1 chokepoint intact rather than forking a second GET path to measure with.

pub async fn undeclare(self) -> Result<()>

Undeclare, telling the network to drop the routing state. The crate’s idiom: teardown is explicit and awaited, never left to Drop.

pub async fn matching_status(&self) -> Result<bool>

Whether any queryable currently matches this querier — “someone serves what we ask”, a routing fact about the querier this process declared (RFC 12 §9’s allowed half). false is not a fleet verdict: it never means “nobody serves this key” (RFC 05 §3.1).

pub async fn matching_events(&self) -> Result<MatchingEvents>

Event-driven matching changes for this querier — same honesty bounds as matching_status.