Struct FeedbackDrivenSelectivity 

pub struct FeedbackDrivenSelectivity { /* private fields */ }

Feedback-driven selective correction

Dynamically adjust the selective estimates based on historical feedback from executions. Uses an exponentially weighted moving average (EWMA) algorithm.

Example

use linkrs_query::optimizer::stats::feedback::selectivity::FeedbackDrivenSelectivity;

let mut feedback = FeedbackDrivenSelectivity::new(0.1);
assert_eq!(feedback.estimated_selectivity(), 0.1);

// Update the feedback
feedback.update_with_feedback(0.15);
feedback.update_with_feedback(0.12);

// Obtain the corrected selective information
let corrected = feedback.corrected_selectivity();
assert!(corrected > 0.0 && corrected <= 0.99);

Implementations

impl FeedbackDrivenSelectivity

pub fn new(estimated_selectivity: f64) -> Self

Create a new feedback-driven, selective estimation method

pub fn with_params( estimated_selectivity: f64, alpha: f64, min_correction: f64, max_correction: f64, ) -> Self

Create using custom parameters

pub fn estimated_selectivity(&self) -> f64

Obtain the original estimate for selectivity.

pub fn set_estimated_selectivity(&mut self, estimated: f64)

Track a fresh uncorrected estimate while preserving the learned correction factor and feedback history.

pub fn corrected_selectivity(&self) -> f64

Obtain the corrected version of the selective content.

pub fn corrected_for_estimate(&self, estimated: f64) -> f64

Correct a fresh estimate with the learned factor without touching the stored baseline, so snapshot reads stay free of planning-order drift.

pub fn correction_factor(&self) -> f64

Obtain the correction factor

pub fn feedback_count(&self) -> u64

Number of times feedback was obtained

pub fn total_input_rows(&self) -> u64

Total measured input rows across applied samples.

pub fn estimation_confidence(&self) -> f64

Obtaining the confidence level for the selected estimate

The confidence level is calculated based on the number of feedbacks received; the more feedback, the higher the confidence level. Return value range: 0.0 - 1.0

pub fn avg_estimation_error(&self) -> f64

Obtain the average estimated error.

pub fn error_std_dev(&self) -> f64

Obtain the standard deviation of the estimated error

pub fn update_with_feedback(&mut self, actual_selectivity: f64)

Update the correction factor (based on the new feedback).

Uses an exponentially weighted moving average (EWMA) algorithm.

pub fn update_with_ratio(&mut self, ratio: f64)

Update the correction factor from an observed estimated-vs-actual row ratio (actual_rows / estimated_rows).

Unlike [update_with_feedback], which receives an absolute selectivity, this receives the correction ratio directly so the feedback loop does not need to know the operator’s input cardinality.

pub fn update_with_measured_ratio(&mut self, ratio: f64, input_rows: u64)

Update the correction factor from a measured local-parameter ratio over a known number of input rows.

This is the feedback loop’s write path. Unlike [update_with_ratio], the ratio must already isolate the operator-local parameter (e.g. observed-vs-estimated selectivity, not output rows), because output ratios conflate upstream cardinality error. Two guards keep distorted measurements out of the factor:

• samples deviating from the learned factor by more than [OUTLIER_DEVIATION] are staged until an independent execution corroborates them (the first sample has no baseline and applies);
• the EWMA step scales with measured input rows, so tiny samples move the factor conservatively.

pub fn update_with_batch(&mut self, actual_selectivities: &[f64])

Batch update of feedback

pub fn reset_correction(&mut self)

Reset the correction factor

pub fn set_alpha(&mut self, alpha: f64)

Setting the EWMA smoothing factor

pub fn set_correction_range(&mut self, min: f64, max: f64)

Set the correction range

pub fn set_selectivity_cap(&mut self, cap: f64)

Setting a selective upper limit

Trait Implementations

impl Clone for FeedbackDrivenSelectivity

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 FeedbackDrivenSelectivity

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

Formats the value using the given formatter.

impl Default for FeedbackDrivenSelectivity

fn default() -> Self

Returns the “default value” for a type.