Suggestion to make it easier to trust Deno's Node compatibility

[ffMathy]

While officially you say that there's Node compatibility, some things are not implemented.

That's fine to me, but it makes it hard for me to decide on whether Deno is a good idea for us.

I'd love to convert our whole codebase, but I am worried that I'll be stuck really late in the process due to incompatibility with some less known library that I need.

I gave Bun a shot. Was really surprised how much is missing compatibility-wise. A colleague even had a Blue Screen of Death while installing packages. They have a similar marketing strategy where they advertise compatibility.

I think this fear could be addressed, and that it could be huge for Deno. I really believe in Deno long term, and I want it to succeed.

My suggestion

What if you started openly working on getting Node's test suite to pass using Deno?

You could even publish a page nightly with GitHub Pages or something similar, which could show the percentage of tests passing, so someone could track Node compatibility over time in a more accurate and confident manner.

Right now you do indeed display which features of which module isn't supported, but what's supported isn't necessarily one-to-one with what actually works. I used many Bun APIs for instance that they said were supported, but due to bugs, it ended up not actually working with particular libraries etc.

Edit: I made this project now, which displays compatibility daily: https://github.com/ffMathy/is-deno-compatible-yet

[bartlomieju]

Hey, thanks for raising this.

That's fine to me, but it makes it hard for me to decide on whether Deno is a good idea for us.

You're not the only one and we are planning to address this more concretely this quarter.

What if you started openly working on getting Node's test suite to pass using Deno?

This has been well underway - we are actually running some tests from Node.js test suite - you can see it here: https://github.com/denoland/deno/blob/main/tests/node_compat/config.jsonc. Admittedly, it's barely ~20% of all tests available in Node and we have some plans to improve it - mainly by running these tests out-of-band, in a separate repository so we can always run all tests and report which ones are actually broken and which ones are now passing. And obviously a frontend page will be added as well to keep track of the status based on Deno release.

Right now you do indeed display which features of which module isn't supported, but what's supported isn't necessarily one-to-one with what actually works. I used many Bun APIs for instance that they said were supported, but due to bugs, it ended up not actually working with particular libraries etc.

We also have another project called npm_smoke_tests that runs various popular npm libraries under Deno and tests if they are working properly. This one is not currently public as it needs some more polish, but then again the plan is to make it open source and widely available.

[kt3k]

https://ffmathy.github.io/is-deno-compatible-yet/

This page seems having only 645 filenames. Where do these come from?

We use this script for copying Node.js test cases from their repository, and that script detects 3681 files under test/ directory of Node.js repo.

This file lists what are remaining for us.

Note: we ignore these directories, and include all others.

[ffMathy]

I made big changes now. It should be calculating correctly from what I can see. I also made it so that you can show the individual tests and their state or collapse them.

It currently shows 7% compatibility. Is that correct now?

[kt3k]

Your list doesn't exclude fixtures, for example. The check here needs to be something like:

!category || ignoredDirectories.some(dir => category.startsWith(dir))

because categories includes items like fixtures/wpt/url, fixtures/es-modules/json-cache, etc. On the other hand, ignoredDirectories list only includes fixtures.

Our counting is visible here: https://github.com/denoland/deno/blob/1d49b3cb0f54eb8184acc00ec4bb3bd519653441/tests/node_compat/runner/TODO.md

In our current calculation, there are 3681 files in total and 595 files are ported for now.

[ffMathy]

Made this PR: #27189

With that merged, I would be able to truly piggyback on the same calculations as you are using.

[ffMathy]

Thank you for merging! I have now corrected the code to use the Deno code. I also converted the codebase from Bun to Deno 2 😄

[nhrones]

Wonderful idea! Eventually could be modeled after the https://caniuse.com/ page.

[ffMathy]

I made this: https://ffmathy.github.io/is-deno-compatible-yet/

But PRs are welcome. It's open source. It updates nightly.

[axetroy]

Good point.

Users need to know how compatible Deno is with Node.

If there is a clear indicator showing more than 95% compatibility, it can greatly increase user confidence and motivate them to migrate to Deno.

[ffMathy]

I made this now: https://ffmathy.github.io/is-deno-compatible-yet/