Struct JobQueue 

pub struct JobQueue { /* private fields */ }

The enqueue side of the queue, plus the runner’s wake-up.

Cloneable and cheap to hold: everything inside is an Arc or a scalar. A subsystem that queues work holds one of these and never sees the runner.

Implementations

impl JobQueue

pub fn new(database: Arc<Database>, config: &JobsConfig) -> Self

pub fn set_max_attempts(&self, max_attempts: u32)

Republishes jobs.max_attempts, for every clone of this queue at once.

Synchronous and infallible, which is what lets a reload call it from cli::apply_reload’s publishing run beside the watch sends.

It sets the budget for work queued from now on and does not touch the backlog: max_attempts is frozen onto each row at enqueue, so a job already waiting keeps the budget it was queued under. Raising this to rescue rows that are about to give up is therefore not what it does — that would be an UPDATE over pending rows, and a deliberately different promise from the one crate::sqlite::job makes.

pub async fn enqueue(&self, spec: JobSpec) -> Result<bool, Error>

Queues one job.

Ok(false) means a live job already holds this (kind, key) — not an error, but the answer two racing callers must both survive. The caller reads it the way RelaySigner::issue reads UpstreamOrder::create’s Ok(None): somebody else is already on it.

pub async fn enqueue_or_log(&self, spec: JobSpec) -> bool

JobQueue::enqueue, logging rather than returning a database failure.

For the callers with nowhere to report one — recover, which the trait defines as best-effort, and any path where refusing to queue is worse than the error it would surface.

pub fn database(&self) -> &Arc<Database>

The database this queue writes to, for a handler that needs one and would otherwise have to be handed a second copy.

Trait Implementations

impl Clone for JobQueue

fn clone(&self) -> JobQueue

Returns a duplicate of the value.

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

Performs copy-assignment from source.