Struct OperatorFeedback 

pub struct OperatorFeedback {
    pub operator_id: String,
    pub operator_type: String,
    pub estimated_rows: u64,
    pub actual_rows: u64,
    pub actual_input_rows: Option<u64>,
    pub partition_count: u32,
    pub exhausted: bool,
    pub estimated_time_us: u64,
    pub actual_time_us: u64,
    pub execution_loops: u64,
    pub condition_key: Option<String>,
    pub shape_key: Option<String>,
}

Operator execution feedback

Records statistical information about the execution of individual operators. Refer to the output format of EXPLAIN ANALYZE in PostgreSQL.

Field Description

operator_id: The unique identifier for the operator. operator_type: The type of the operator (such as Scan, Filter, Join, etc.) estimated_rows: The number of output rows estimated by the optimizer. actual_rows: The actual number of rows output. estimated_time_us: Estimated execution time (in microseconds) actual_time_us: Actual execution time (in microseconds) execution_loops: The number of executions (for example, the number of times the inner loop of a Nested Loop is executed).

Example

use linkrs_query::optimizer::stats::feedback::query::OperatorFeedback;

let feedback = OperatorFeedback {
    operator_id: "scan_1".to_string(),
    operator_type: "IndexScan".to_string(),
    estimated_rows: 100,
    actual_rows: 150,
    actual_input_rows: None,
    partition_count: 1,
    exhausted: true,
    estimated_time_us: 1000,
    actual_time_us: 1200,
    execution_loops: 1,
    condition_key: None,
    shape_key: None,
};

assert_eq!(feedback.row_estimation_error(), 0.5); // (150-100)/100

Fields

operator_id: String

Operator ID

operator_type: String

Operator types

estimated_rows: u64

Estimated number of output lines

actual_rows: u64

Number of actual lines of output

actual_input_rows: Option<u64>

Actual input rows consumed by this operator, summed across partitions. Lets the fold loop isolate the operator-local parameter (e.g. observed selectivity) from upstream cardinality error. None means unmeasured: the sample cannot separate predicate error and must not train selectivity.

partition_count: u32

How many partition instances this sample aggregates. Always >= 1; greater than 1 means per-partition profiles were summed into one per-execution sample before any correction was computed.

exhausted: bool

Whether every aggregated instance ran to exhaustion. Samples from starved operators (Limit / early termination) carry partial counts and must not train the feedback loop.

estimated_time_us: u64

Estimated execution time (in microseconds)

actual_time_us: u64

Actual execution time (in microseconds)

execution_loops: u64

Number of executions (for example, the number of times the inner loop of a nested loop is executed)

condition_key: Option<String>

Normalized predicate key for filter operators.

Populated only for operators whose spec carries a predicate (e.g. UnarySpec::Filter); used by the statistics feedback loop to attach estimated-vs-actual row ratios to a specific predicate.

shape_key: Option<String>

Normalized operator shape key ("{space}:{Type}:{discriminator}").

Populated for operators whose output cardinality is estimated independently (scans, graph traversals, joins, applies); used by the cardinality feedback loop to correct per-shape row estimates. Filter operators are keyed by condition_key instead and leave this None.

Implementations

impl OperatorFeedback

pub fn new( operator_id: String, operator_type: String, estimated_rows: u64, actual_rows: u64, ) -> Self

Create new operator feedback.

pub fn row_estimation_error(&self) -> f64

Estimation error of the number of lines calculated

Return the relative error: (Actual – Estimated) / Estimated If the estimate is 0, return 1.0

pub fn time_estimation_error(&self) -> f64

Calculating the estimated error in time measurement

Return relative error: (actual-estimate)/estimate

pub fn avg_rows_per_loop(&self) -> f64

Obtain the average number of actual rows executed in each case.

For operators that are executed multiple times (such as the inner loop in a Nested Loop), Return the average number of lines executed per run.

pub fn avg_time_us_per_loop(&self) -> f64

Obtain the average actual time (in microseconds) for each execution.

Trait Implementations

impl Clone for OperatorFeedback

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for OperatorFeedback

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

Formats the value using the given formatter.