Struct Executor 

pub struct Executor { /* private fields */ }

Concurrent scheduler-dispatch executor.

Implementations

impl Executor

pub fn new() -> Self

pub fn with_config(config: ExecutorConfig) -> Self

pub fn register_actor( &self, root_id: ProjectRootId, ctx: Arc<AppContext>, ) -> bool

Register an actor if one is not already present.

Existing actors keep their current context and scheduler state; subc routing reuses them and reconfigures through the Mutating lane rather than replacing the per-root AppContext. Returns true when a new actor was inserted.

pub fn remove_actor(&self, root_id: &ProjectRootId)

Remove an actor from scheduler state.

This is intentionally minimal: subc uses it only for a just-created RouteBind actor whose configure failed before any route was installed, so there is no in-flight work to quiesce. The removed AppContext is dropped after releasing the scheduler lock so watcher/LSP teardown never runs under that mutex.

pub fn cancel_queued_maintenance(&self, root_id: &ProjectRootId) -> usize

Cancel maintenance jobs that have not started for one retained actor.

Interactive work and already-running maintenance remain untouched. Each cancelled job receives a normal completion so its caller can settle bookkeeping through the same path as an executed job.

pub fn actor_registered(&self, root_id: &ProjectRootId) -> bool

Return whether scheduler state currently has an actor for this root.

pub fn actor_context(&self, root_id: &ProjectRootId) -> Option<Arc<AppContext>>

Snapshot one actor context without retaining the scheduler lock while maintenance drops root-scoped resources.

pub fn actor_contexts(&self) -> Vec<Arc<AppContext>>

Snapshot the registered actor contexts.

The returned Arcs keep contexts alive after the scheduler lock is released, so callers can run teardown without holding executor state.

pub fn actor_entries(&self) -> Vec<(ProjectRootId, Arc<AppContext>)>

Snapshot the registered root ids paired with their actor contexts.

pub fn try_actor_entries(&self) -> Option<Vec<(ProjectRootId, Arc<AppContext>)>>

Non-blocking variant for the health path: the probe reply must stay cheap under any load, so it skips the actor list (reported as busy) rather than waiting on the scheduler state lock.

pub fn submit( &self, root_id: ProjectRootId, lane: Lane, request_id: String, job: ExecutorJob, ) -> CompletionHandle

pub fn submit_async( &self, root_id: ProjectRootId, lane: Lane, request_id: String, job: ExecutorJob, ) -> Receiver<Response>

pub fn submit_cancellable_async( &self, root_id: ProjectRootId, lane: Lane, request_id: String, job: ExecutorJob, ) -> (Receiver<Response>, JobCancellation)

Submit an interactive job with an exact-job cancellation token.

The returned token cancels THIS job only (queued: removed and settled with request_cancelled; running: signalled cooperatively). The job observes the token via current_job_cancellation.

pub fn cancel_job( &self, root_id: &ProjectRootId, token: &JobCancellation, ) -> JobCancelOutcome

Cancel one exact job by its token.

Queued jobs are removed and settled immediately; running jobs are signalled and return through their normal completion path at the next cooperative checkpoint. Jobs that sealed their commit finish normally.

pub fn submit_maintenance_async( &self, root_id: ProjectRootId, lane: Lane, request_id: String, job: ExecutorJob, ) -> Receiver<Response>

pub fn pool_size(&self) -> usize

pub fn actor_cap(&self) -> usize

pub fn read_cap(&self) -> usize

pub fn heavy_permits(&self) -> usize

pub fn interactive_reserve(&self) -> usize

pub fn maintenance_cap(&self) -> usize

pub fn try_dispatch_liveness_snapshot(&self) -> Option<DispatchLivenessSnapshot>

pub fn try_mutating_lane_snapshots(&self) -> Option<Vec<MutatingLaneSnapshot>>

pub fn try_mutating_job_state_label( &self, root_id: &ProjectRootId, request_id: &str, ) -> Option<&'static str>

pub fn try_bind_blocker_snapshot( &self, root_id: &ProjectRootId, request_id: &str, ) -> Option<BindBlockerSnapshot>

Snapshot RouteBind blockers without waiting on scheduler state. The subc health path uses this only for a delayed-bind breadcrumb, so contention is reported as scheduler busy rather than delaying the transport loop.

pub fn nonrunnable_dispatch_count(&self) -> usize

pub fn completion_counts(&self) -> (u64, u64)

pub fn actor_is_fatal(&self, root_id: &ProjectRootId) -> bool

Trait Implementations

impl Default for Executor

fn default() -> Self

Returns the “default value” for a type.