Struct SharedScheduler 

pub struct SharedScheduler { /* private fields */ }

Engine-level shared scheduler.

Created once at database startup and reused across all queries. Each query receives an Arc reference through its bindings. Workers are created once and persist across query boundaries, eliminating the per-query thread overhead of the old MorselWorkerPool pattern.

Query teardown does not join or shut down the shared workers — only engine shutdown calls SharedScheduler::shutdown.

Implementations

impl SharedScheduler

pub fn new(max_workers: usize) -> Self

Create a new shared scheduler with max_workers persistent threads.

At least one worker is always created. Pass 0 to use the default (typically number of CPU cores, minimum 1).

pub fn apply_to_runtime(&self, runtime: &ExecutionRuntime)

Inject this scheduler into the given runtime.

This makes the runtime’s worker pool point to the shared scheduler’s underlying MorselWorkerPool, so that all Exchange/Gather operators in that query use the shared worker threads.

pub fn max_workers(&self) -> usize

Number of workers in this scheduler.

pub fn shutdown(&mut self)

Shut down all workers and wait for them to exit.

Called during engine shutdown. After this, no further batches can be submitted. Must only be called once no more queries need the scheduler.

pub fn shutdown_shared(&self)

Shut down workers through a shared reference.

Signals the stop flag and closes the batch channel so workers exit promptly, even while other references (in-flight query runtimes) still exist. Worker threads are not joined here; they terminate after observing the stop flag / closed channel.

Trait Implementations

impl Debug for SharedScheduler

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

Formats the value using the given formatter.