Struct SelectivityEstimator 

pub struct SelectivityEstimator { /* private fields */ }

Selective Estimator

Provide selective estimates based on statistical information and heuristic rules.

Implementations

impl SelectivityEstimator

pub fn new(stats_manager: Arc<StatisticsManager>) -> Self

Create a new selective estimator.

pub fn with_feedback( stats_manager: Arc<StatisticsManager>, feedback: Arc<SelectivityFeedbackManager>, ) -> Self

Create a selective estimator wired to a feedback correction manager.

When feedback is enabled, estimate_from_expression consults the corrected selectivity for the normalized condition key first and registers newly estimated conditions for later correction.

pub fn feedback_manager(&self) -> Option<&Arc<SelectivityFeedbackManager>>

Return the underlying feedback correction manager, if any.

pub fn snapshot(&self) -> SelectivitySnapshot

Capture the learned corrections for one planning decision to read through. Empty when no feedback manager is attached.

pub fn estimate_readonly( &self, space: Option<&str>, expr: &Expression, tag_name: Option<&str>, snapshot: &SelectivitySnapshot, ) -> f64

Estimate a predicate without touching shared feedback state.

Unlike estimate_from_expression, this neither registers the condition nor refreshes its baseline: the snapshotted correction (if any) applies to the fresh estimate, so repeated calls stay identical until execution feedback lands.

pub fn estimate_equality_selectivity( &self, space: Option<&str>, tag_name: Option<&str>, property_name: &str, value: Option<&Value>, ) -> f64

Estimated Equivalence Condition Selectivity

If there is histogram statistical information, use the histogram for an accurate estimation. Otherwise, if basic statistical information is available, use 1 divided by the number of distinct values. Otherwise the default value of 0.1 is used

pub fn estimate_equality_selectivity_simple( &self, space: Option<&str>, tag_name: Option<&str>, property_name: &str, ) -> f64

Estimated Equivalence Condition Selectivity (simplified version, without values)

pub fn estimate_range_selectivity( &self, space: Option<&str>, tag_name: Option<&str>, property_name: &str, range: &RangeCondition, ) -> f64

Conditional selectivity of the estimation range

If there is histogram statistical information, the calculations are based on that histogram. Otherwise, use the default value of 1/3.

pub fn estimate_range_selectivity_with_bounds( &self, min_val: f64, max_val: f64, range_size: f64, ) -> f64

Conditional Selectivity of Estimation Ranges (with Boundary Values)

Adjust the selectivity based on the size of the scope.

pub fn estimate_ordered_comparison( &self, space: Option<&str>, tag_name: Option<&str>, property_name: &str, bound: Option<f64>, greater_than: bool, ) -> f64

Range-comparison selectivity from the sampled min/max envelope.

Linear interpolation over [min, max] assuming uniform distribution: prop > bound covers (max - bound) / (max - min) of the range. Falls back to the fixed heuristics when no envelope is available.

pub fn estimate_less_than_selectivity(&self, value: Option<f64>) -> f64

Estimated to be less than conditional selectivity

If there is statistical information available, the calculations are based on histograms. Otherwise, assume that the data is evenly distributed, and return 1/3.

pub fn estimate_greater_than_selectivity(&self, value: Option<f64>) -> f64

The estimate is greater than the condition selectivity.

pub fn estimate_like_selectivity(&self, pattern: &str) -> f64

Estimating the selectivity of LIKE conditions

Adjust the selectivity based on the prefix and suffix wildcards of the pattern:

• prefix%: high selectivity (about 0.1)
• %suffix: medium selectivity (about 0.2)
• %substring%: less selective (about 0.5)
• No wildcard: exact match (about 0.05)

pub fn estimate_in_selectivity(&self, list_size: usize) -> f64

Estimating the selectivity of the IN list

Assuming that the selectivity of each value is the same, the total selectivity = number of values * selectivity of a single value.

pub fn estimate_is_null_selectivity(&self) -> f64

Estimated IS NULL selectivity

pub fn estimate_is_not_null_selectivity(&self) -> f64

The “IS NOT NULL” condition is used for selection purposes.

pub fn estimate_not_selectivity(&self, inner_selectivity: f64) -> f64

Estimated NOT conditional selectivity

The selectivity of the NOT condition = 1 – the selectivity of the original condition

pub fn estimate_from_expression( &self, space: Option<&str>, expr: &Expression, tag_name: Option<&str>, ) -> f64

Estimating selectivity from expressions

This is the main entry method, which distributes the data to the specific estimation methods based on the type of the expression.

When a feedback correction manager is attached, the normalized condition key is looked up first: a registered correction overrides the histogram / heuristic estimate. Otherwise the estimate is computed and registered so later executions can correct it.

Trait Implementations

impl Clone for SelectivityEstimator

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 SelectivityEstimator

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

Formats the value using the given formatter.