Crate perfgate_budget 

Budget evaluation logic for performance thresholds.

This crate provides pure budget evaluation functions with no I/O dependencies. It handles threshold checking, regression calculation, and verdict aggregation.

Part of the perfgate workspace.

Overview

The crate provides:

• evaluate_budget - Evaluate a single metric against a budget threshold
• calculate_regression - Calculate regression percentage between baseline and current
• determine_status - Determine metric status (Pass/Warn/Fail) from regression
• aggregate_verdict - Aggregate multiple metric statuses into a final verdict

Example

use perfgate_budget::{evaluate_budget, calculate_regression, determine_status};
use perfgate_types::{Budget, Direction, MetricStatus};
use std::collections::BTreeMap;

// Create a budget for a lower-is-better metric (e.g., wall time)
let budget = Budget {
    noise_threshold: None,
    noise_policy: perfgate_types::NoisePolicy::Ignore,
    threshold: 0.20,       // 20% regression fails
    warn_threshold: 0.10,  // 10% regression warns
    direction: Direction::Lower,
};

// Evaluate baseline vs current
let baseline = 100.0;
let current = 115.0;

let result = evaluate_budget(baseline, current, &budget, None).unwrap();

assert_eq!(result.status, MetricStatus::Warn);
assert!((result.regression - 0.15).abs() < 1e-10);

Structs

BudgetResult

Result of evaluating a single metric against a budget.

Enums

BudgetError

Errors that can occur during budget evaluation.

Functions

aggregate_verdict

Aggregates multiple metric statuses into a final verdict.

calculate_regression

Calculates the regression percentage between baseline and current values.

determine_status

Determines the metric status based on regression and thresholds.

evaluate_budget

Evaluates a metric against a budget threshold.

evaluate_budgets

Evaluates multiple metrics against their budgets.

reason_token

Generates a reason token for a metric status.