Write benchmarks with Ende.bench

Ende has a built-in benchmark runner. Register benchmarks with Ende.bench and run them with the ende bench subcommand, which reports timings and compares grouped benchmarks against a baseline.

Edit on GitHub

A classic question: for membership checks, how much faster is a Set than an Array? Prepare both with the same ten thousand items, then look for an element near the end, the worst case for the array.

const items = Array.from({ length: 10_000 }, (_, i) => `item-${i}`);
const itemsArray = [...items];
const itemsSet = new Set(items);
const needle = "item-9999";

Benchmarks that belong to the same group are compared with each other in the summary. Exactly one of them can be marked as the baseline.

Ende.bench({
  baseline: true,
  fn: () => {
    itemsArray.includes(needle);
  },
  group: "lookup",
  name: "Array.prototype.includes"
});

Ende.bench({
  fn: () => {
    itemsSet.has(needle);
  },
  group: "lookup",
  name: "Set.prototype.has"
});

Sometimes a benchmark needs setup that should not be measured. The bench context passed to the function exposes start and end methods that limit the measured section. Here only the join call is timed, not building the input array.

Ende.bench("join parts", (b) => {
  const parts = Array.from({ length: 1_000 }, (_, i) => `part-${i}`);
  b.start();
  parts.join(", ");
  b.end();
});

Run the file with ende bench. The trailing columns with min, max and percentiles are trimmed here to keep the table narrow: ende bench ende_bench.ts CPU | Apple M1 Max Runtime | Ende 2026.10.17 (aarch64-apple-darwin) | benchmark | time/iter (avg) | iter/s | | -------------------------- | --------------- | ------------- | | join parts | 15.8 µs | 63,360 | group lookup | Array.prototype.includes | 9.9 µs | 101,400 | | Set.prototype.has | 4.3 ns | 233,500,000 | summary Array.prototype.includes 2303x slower than Set.prototype.has The linear scan takes microseconds while the hash lookup takes nanoseconds, so the Set wins by more than three orders of magnitude.

Useful flags: ende bench --filter lookup runs only matching benchmarks, and ende bench --json emits machine-readable results.

Run this example locally using the Ende CLI:

ende bench https://docs.ende.sh/examples/scripts/write_benchmarks.ts

Write benchmarks with Ende.bench

Additional resources