Struct ABTestEvaluator 

pub struct ABTestEvaluator { /* private fields */ }

Evaluator that collects prediction pairs and judges the test.

Thread-safe — mutable state is wrapped in Mutex.

Implementations

impl ABTestEvaluator

pub fn new(config: ABTestConfig) -> Self

Creates a new A/B test evaluator with the given configuration.

pub fn should_use_challenger(&self, request_id: u64) -> bool

Decide whether a specific request should be routed to the challenger.

Uses a simple deterministic hash on the request_id for consistent routing (no randomness in tight loops).

pub fn log_prediction( &self, actual: f64, champion_pred: f64, challenger_pred: f64, )

Log a prediction pair (both champion and challenger predictions along with the actual observed value).

pub fn evaluate(&self) -> ABTestResult

Evaluate the A/B test based on logged records.

Computes metrics over the most recent evaluation_windows records to avoid stale early predictions diluting accuracy.

pub fn champion_metrics(&self) -> Option<VersionAccuracy>

Champion accuracy from logged data.

pub fn challenger_metrics(&self) -> Option<VersionAccuracy>

Challenger accuracy from logged data.

pub fn record_count(&self) -> usize

Number of logged records.

Trait Implementations

impl Debug for ABTestEvaluator

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

Formats the value using the given formatter.