It’s alive!

I tried a few of the modules listed, to find some good stuff and make tests.

@libs/markdown: import * as mod from "jsr:@libs/markdown"; crashed with Error: Missing "./slugify" specifier in "@jsr/std__text" package. Did not find how to fix this one.

@ch/tomd: import { tomd } from "jsr:@ch/tomd" displays SyntaxError: The requested module '/_npm/[email protected]/_esm.js' does not provide an export named 'Browser'

I managed to crash the preview server when I tried to import k-means-pp from jsr:k-means-pp (the correct specifier was jsr:@ppz/k-means-pp), my mistake.

(Maybe these errors can be controlled?)

Most modules I've tried worked out the the box. There is good stuff: @ppz/k-means-pp — @storynode/jql — @dbushell/hyperless — @notwoods/spherical-geometry-js…

I'm watching https://www.youtube.com/watch?v=MFCn4ce5dVc for the rationale.

Not sure if we will be able to fix any of those errors, but I can look. Do you think they are blocking?

No, not blocking. I'm now trying to publish a package to see what happens "end-to-end".

The whole experience seems very nice.

The slugify error seems like a bug in @libs/markdown. In @std/text 1.0.5, slugify was available but documented as unstable; in @std/text 1.0.6 it was moved to unstable_slugify. But @libs/markdown says it just depends on @std/text 1, but it needs exactly 1.0.3, 1.04, or 1.0.5 to work. It has 1.0.4 in its deno.lock file but I don’t think we have to (or should) respect lock files when resolving transitive dependencies.

The @ch/tomd error is because we’re trying to load the happy-dom package from jsDelivr, but jsDelivr refuses to serve it and replaces it with an error saying:

Failed to bundle using Rollup v2.79.1: the file imports a not supported node.js built-in module "node:perf_hooks". If you believe this to be an issue with jsDelivr, and not with the package itself, please open an issue at https://github.com/jsdelivr/jsdelivr'

I’m not entirely sure why jsDelivr’s error isn’t shown; but at any rate the error saying that it doesn’t export a symbol called Browser is accurate, because jsDelivr wasn’t able to bundle the code. I don’t think we need to do anything for this one either; this is just a package that doesn’t work in a browser environment. (Though perhaps it would be nice to detect when jsDelivr fails to bundle the code, rather than downloading a broken library… I can file a bug for that.)

I fixed the jsr:k-means-pp error, and also am now populating the in-memory JSR version resolution cache correctly. There are probably some more bugs to fix here. We need more tests.