benchmark.ts

// Throughput of the core operations. Run with: npx tsx benchmark/benchmark.ts
//
// Reports how fast the library fits a posterior and produces forecasts, so the
// cost of running predictive maintenance on a large fleet is clear.

import {
  fitExponential,
  fitWeibull,
  forecast,
  mulberry32,
  sampleWeibull,
} from "../src/index.js";
import type { SurvivalObservation } from "../src/index.js";

const DATASET_SIZE = 2000;
const FIT_ITERATIONS = 200;
const FORECAST_ITERATIONS = 2000;

function simulate(seed: number): SurvivalObservation[] {
  const rnd = mulberry32(seed);
  return Array.from({ length: DATASET_SIZE }, () => ({ duration: sampleWeibull(rnd, 2, 50), censored: false }));
}

function timed(label: string, iterations: number, run: (index: number) => void): void {
  const start = performance.now();
  for (let i = 0; i < iterations; i += 1) run(i);
  const elapsedMs = performance.now() - start;
  const perSecond = (iterations / elapsedMs) * 1000;
  console.log(
    `${label}: ${iterations} iterations in ${elapsedMs.toFixed(1)} ms ` +
      `(${perSecond.toFixed(0)} per second, ${(elapsedMs / iterations).toFixed(3)} ms each)`,
  );
}

const dataset = simulate(1);

console.log(`Dataset: ${DATASET_SIZE} observations\n`);

timed("Exponential fit", FIT_ITERATIONS, () => {
  fitExponential(dataset);
});

timed("Weibull fit (60-point shape grid)", FIT_ITERATIONS, () => {
  fitWeibull(dataset);
});

const weibullModel = fitWeibull(dataset);
timed("Weibull forecast (3 horizons)", FORECAST_ITERATIONS, (i) => {
  forecast(weibullModel, { age: i % 50, horizons: [6, 24, 72] });
});