underscore-benchmark

Benchmarking stats calculation from offen/offen against different underscore versions.

Default setup

To run the benchmark from this repository, you first need to install the project's dependencies. Next, you can run the setup which does the following:

• pull the underscore versions to compare and put them in the vendor directory
• create fixture data for the benchmark to run against

npm install
npm run setup

Once this has finished, you are ready to run the benchmark in Node.js:

npm test

You can also run the benchmark in a browser:

npm start

budo will start a local server at port 9966: loading this URL in a browser will run the benchmark. Results will be printed to your browser's console.

---

If used with defaults, the benchmark will use commit c9b4b63fd08847281260205b995ae644f6f2f4d2 as the baseline and eaba5b58fa8fd788a5be1cf3b66e81f8293f70f9 as the comparison for the benchmark.

Adjusting the benchmark

Running against a single version only

If you do not want to run the benchmark against both the baseline and the comparison version at the same time, you can choose to run a single test only.

Using the CLI, this is done by passing the desired test as the first command line arg:

npm t -- baseline

If you are using the browser, append the name of the desired case as the URL hash and reload the page, e.g. http://localhost:9966/#baseline

Running against different versions of underscore

If you want to run this benchmark against different versions of underscore, you can use the ./pull.sh script to pull any versions you want to compare from the jashkenas/underscore repository.

The script expects two arguments, the baseline ref and the comparison ref:

node ./scripts/pull-versions.js master some-feature-branch

Such an argument can be any either a branchname or a commit hash for the repository.

Running against more/less user data

By default, the benchmark runs against randomly created user data for 5000 users. This is an average real-world workload for the code under test, so it's probably a sane default. If you want to benchmark how two versions of underscore compare with more or less data being processed, you can re-run the fixture creation, passing the non-default number of users as the first argument:

node ./scripts/generate-fixtures.js 2500 ./fixtures/events.json

Reverting your adjustments

If you want to revert your benchmark setup to the default state after making adjustments, you can use the setup command:

npm run setup