[Idea]: use ES6 modules for running unit tests and benchmarks in web browsers

[kgryte]

Idea

Currently, when generating stdlib API documentation, we generate UMD bundles for unit tests and benchmarks. When a user navigates to our package documentation, they can load unit tests and benchmarks and have those run without needing to setup a local environment. The pain point here is that creating separate bundles for each package is time consuming and adds significant heft to the www repo.

The goal of this idea is to refactor the way we support unit tests and benchmarks to use ES6 modules and potentially skip bundling altogether. This has the downside of not supporting older browsers which don't support the <module> tag, but is probably fine considering that running package unit tests and benchmarks is likely a forward looking concern.

Expected Outcomes

Users will be able to run unit tests and benchmarks directly in their web browsers by navigating to project API documentation and what is loaded are ES6 modules, not UMD bundles.

Involved Software

No other software is necessary.

Prerequisite Knowledge

JavaScript, Node.js, HTML/CSS, JSX/React.

Difficulty

Intermediate.

Project Length

350 hours.

Potential Mentors

@kgryte @Planeshifter @steff456

[kgryte]

@Planeshifter Now that we have UMD and ES Module builds in standalone repositories, this could be more straightforward, I believe, as we may be able to load directly using JSDelivr rather than the stdlib hosted server. What are your thoughts?

[Planeshifter]

@kgryte Yes, should be fairly straightforward to load the ES modules from JSDelivr.
This project may be helped by writing own test runner as planned (#7).

[kgryte]

Good point. We'd need to account for 3rd party dev dependencies (e.g., tape, proxyquire, and the like).

[kgryte]

Marking this idea as higher priority, as the ability to run unit tests, examples, and benchmarks via ES modules could potentially drastically improve website build times and pave the way for nightly website builds.

If someone is interested in pursuing this idea, they are advised to spend time reading the current website code located in the stdlib-js/www repository. Your proposal should reflect our current technology stack, and you should describe the concrete steps necessary to achieve expected outcomes.

[kgryte]

@Planeshifter We initially marked this as 350 hours. Given the prior efforts to create ES Module builds, is this an appropriate project length or could the expected outcomes be achieved in 90 or 175 hours?