austindd / rescript-benchmarkjs

ReScript bindings to the Benchmark.js library

rescript-benchmarkjs

ReScript bindings to Benchmark.js, an excellent performance benchmarking library.

Installation

Using npm:

npm install --save-dev rescript-benchmarkjs

Using yarn:

yarn add --dev rescript-benchmarkjs

Don't forget to add the dependency to your bsconfig.json file:

{
  "bs-dev-dependencies": [
    "rescript-benchmarkjs"
  ]
}

Usage

open BenchmarkJs

let suiteConfig = {
  ...Suite.defaultConfig,
  onStart: _ => Js.log("-- Running Benchmark Suite --\r\n"),
  onComplete: _ => Js.log("-- Benchmark Results --\r\n"),
}

let benchmarkConfig = {
  ...Benchmark.defaultConfig,
  onError: Js.log,
  onStart: event => Js.log("Running: '" ++ Benchmark.name(event.target) ++ "'..."),
  onComplete: _ => Js.log("Done\r\n"),
}

Random.init(486)
let myArray = Belt.Array.makeBy(1000, _ => Random.int(99999))

Suite.make("Array Sort", ~config=suiteConfig)
->Suite.add("Belt.SortArray.stableSortBy", ~config=benchmarkConfig, (. ()) => {
  let result = myArray->Belt.SortArray.stableSortBy(Pervasives.compare)
  ignore(result)
})
->Suite.add("Js.Array2.slice + Js.Array2.sortInPlace", ~config=benchmarkConfig, (. ()) => {
  let result = myArray->Js.Array2.sliceFrom(0)->Js.Array2.sortInPlace
  ignore(result)
})
->Suite.run
->Suite.toArray
->Belt.Array.map(Benchmark.toString)
->Belt.Array.forEach(Js.log)

Important Notes:

1. One obvious difference bewteen the ReScript interface and the original Benchmark.js API is that, in ReScript, we cannot define "setup code" in one function, and then expect the values defined there to be available in a separate function defining a "test case". Therefore, any values you expect to be available within both the setup function AND a test case function must be defined in an outer (usually global) scope. This is done in the example above with the myArray value.
2. It's generally a good idea to inspect the generated code for each test case. In some cases, you may find that the compiler has optimized or de-optimized a test case in unexpected ways. This might include code removal, adding/removing wrapper objects, using helper functions for type-safety, etc. In some cases, you might intend to capture this compiler behavior in your tests. In other cases, you may find yourself running tests with false assumptions about the generated JS code. Always check the generated code to be certain.