Struct QueryManager 

pub struct QueryManager { /* private fields */ }

Query Manager

Single enum dual emission source: query lifecycle halves (QueryStarted / QueryCompleted / SlowQueryDetected) are emitted here, session lifecycle halves (SessionCreated / SessionDestroyed) by GraphSessionManager. Both sides can share one Arc<EventSubscriptions<SessionEvent>> via new_with_shared so observers subscribe once and receive both halves.

Implementations

impl QueryManager

pub fn new() -> Self

pub fn new_with_shared(shared: Arc<EventSubscriptions<SessionEvent>>) -> Self

Build a manager sharing one session-event registry with GraphSessionManager.

pub fn shared_session_callbacks(&self) -> Arc<EventSubscriptions<SessionEvent>>

Shared session-event registry behind this manager.

pub fn set_shared_session_callbacks( &mut self, shared: Arc<EventSubscriptions<SessionEvent>>, )

Point this manager at a shared session-event registry.

pub fn register_session_callback( &self, callback: SessionEventCallback, ) -> SubscriptionId

Register a runtime observer for session/query events.

pub fn register_session_callback_filtered( &self, callback: SessionEventCallback, filter: EventFilter<SessionEvent>, ) -> SubscriptionId

Register a filtered observer invoked only when filter returns true.

pub fn unregister_session_callback(&self, id: SubscriptionId) -> bool

Remove a previously registered observer. Returns true if present.

pub fn session_callback_count(&self) -> usize

Number of registered session observers.

pub fn set_query_registry(&self, registry: Arc<QueryRegistry>)

Attach the executor’s query registry so kill_query also cancels the execution token. Without it, kill_query only marks the query status.

pub fn clear_query_registry(&self)

Detach the executor’s query registry; kill_query returns to status-marking only.

pub fn register_progress_callback( &self, callback: QueryProgressCallback, ) -> SubscriptionId

Register a progress observer. The executor invokes it at its own row interval; unregistered by default, hence zero hot-path overhead.

pub fn register_progress_callback_with_interval( &self, callback: QueryProgressCallback, rows_interval: u64, ) -> SubscriptionId

Register a progress observer thinned to every rows_interval rows.

Observer-side thinning: deliveries whose rows_processed is not a multiple of rows_interval are dropped before reaching callback. rows_interval == 0 disables thinning (every emission is delivered). The executor still decides when to call emit_progress; pairing a coarse executor cadence with a fine observer interval costs nothing extra, while a fine executor cadence with a coarse observer interval only pays for the dropped filter checks.

pub fn register_progress_callback_filtered( &self, callback: QueryProgressCallback, filter: EventFilter<QueryProgress>, ) -> SubscriptionId

Register a filtered progress observer.

pub fn unregister_progress_callback(&self, id: SubscriptionId) -> bool

Remove a previously registered progress observer.

pub fn progress_callback_count(&self) -> usize

Number of registered progress observers.

pub fn emit_progress(&self, session_id: i64, query_id: i64, rows_processed: u64)

Emit a progress notification. No-op when nobody listens.

Explicit executor entry point: streaming operators call this when their processed row count reaches the configured cadence. Callers outside the executor (tests, manual drivers) may invoke it directly.

pub fn set_slow_query_threshold_ms(&self, threshold_ms: i64)

Configure the threshold used to emit SlowQueryDetected.

pub fn slow_query_threshold_ms(&self) -> i64

Current slow-query threshold in milliseconds.

pub fn register_query( &self, session_id: i64, user_name: String, space_name: Option<String>, query_text: String, ) -> i64

Register a new query

pub fn register_query_with_id( &self, query_id: i64, session_id: i64, user_name: String, space_name: Option<String>, query_text: String, ) -> ManagerResult<()>

Register a new query under a caller-assigned id.

The network service assigns one monotonic id per execution and uses it for the engine request, this registry, and the session tracker, so SHOW QUERIES rows double as KILL QUERY targets. The id lane advances past the fixed id to keep generated ids collision-free. An id already held by a running query is rejected; a stale finished entry under the same id is replaced.

pub fn finish_query(&self, query_id: i64) -> ManagerResult<()>

Complete the query.

pub fn fail_query(&self, query_id: i64) -> ManagerResult<()>

The marker query failed.

pub fn kill_query(&self, query_id: i64) -> ManagerResult<()>

Terminate the query.

Marks the query Killed and, when a query registry is attached via set_query_registry, cancels its execution token with CancelReason::UserKill so a running query actually stops. Without an attached registry this only marks the status; the executor-side QueryRegistry remains the source of truth for cancellation. Query ids are generated as positive i64 values, so a negative id is rejected as invalid input instead of wrapping on the i64 to executor u64 id conversion.

pub fn get_query(&self, query_id: i64) -> Option<QueryInfo>

Obtain the query information

pub fn get_all_queries(&self) -> Vec<QueryInfo>

Retrieve all queries

pub fn get_running_queries(&self) -> Vec<QueryInfo>

Obtain the queries that are currently running.

pub fn cleanup_finished_queries(&self, keep_count: usize)

Clean up the completed queries (retaining only the last N of them).

Trait Implementations

impl Debug for QueryManager

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

Formats the value using the given formatter.

impl Default for QueryManager

fn default() -> Self

Returns the “default value” for a type.