#!/usr/bin/env node
/**
 * Bench harness — runs analyzers over seeded corpus fixtures and computes
 * precision, recall, and F1 against labeled ground truth.
 *
 * Spec 11 R2
 *
 * Usage:
 *   pnpm bench              # Run all corpus analyzers
 *   pnpm bench --sweep      # Sweep mode: vary thresholds and report curves
 *   pnpm bench --json       # Output machine-readable JSON only
 */
import type { Violation } from '../types.js';
interface ExpectedEntry {
    file: string;
    rule: string;
    symbol?: string;
    severity?: string;
    reason?: string;
}
interface MatchResult {
    expected: ExpectedEntry;
    matched: boolean;
    actualViolation?: Violation;
}
export interface RuleMetrics {
    rule: string;
    truePositives: number;
    falsePositives: number;
    falseNegatives: number;
    precision: number;
    recall: number;
    f1: number;
    knownMisses: number;
    recoveredMisses: number;
    trueRecall: number;
    trueF1: number;
}
export interface AnalyzerMetrics {
    truePositives: number;
    falsePositives: number;
    falseNegatives: number;
    precision: number;
    /** Effective recall — computed against reconciled ground truth (excl. known misses). Gates regressions. */
    recall: number;
    /** Effective F1 — from precision + effective recall. */
    f1: number;
    /** True recall — computed against full ground truth (incl. known misses). Measures distance from done. */
    trueRecall: number;
    /** True F1 — from precision + true recall. */
    trueF1: number;
    /** Count of ground-truth violations that are known misses (not detected by the analyzer). */
    knownMisses: number;
    /** Count of known-miss entries that WERE detected — a limitation was fixed. */
    recoveredMisses: number;
    /** Per-rule metrics — precision/recall/F1 for each rule within the analyzer. */
    ruleMetrics: RuleMetrics[];
    nearMissResults: Array<{
        file: string;
        violations: number;
        passed: boolean;
    }>;
    details: MatchResult[];
    knownMissDetails: MatchResult[];
    warnings: string[];
    /** Count of violations that violate the hook contract (empty file path or line 0). Must always be zero. */
    hookContractViolations: number;
}
export interface BenchReport {
    timestamp: string;
    summary: {
        totalAnalyzers: number;
        passed: number;
        failed: number;
        totalKnownMisses: number;
        microAvgPrecision: number;
        microAvgRecall: number;
        microAvgF1: number;
        microAvgTrueRecall: number;
        microAvgTrueF1: number;
    };
    analyzers: Record<string, AnalyzerMetrics>;
    baselineComparison?: {
        regressions: string[];
        improvements: string[];
    };
}
/**
 * Run the full bench harness and return the report.
 * Exported so tests can assert bench results programmatically.
 */
export declare function runBench(): Promise<BenchReport>;
export {};
//# sourceMappingURL=runBench.d.ts.map