sbjs

It's a highly optimized and extremely simple yet robust implementation of it, sure. Is that reason to dismiss it?

Consider Vite's node-side HMR implementation. It creates its own module system on top of Node's native module system, using `node:vm`. So its modules are really second class citizens that have to be glued to the native module system.

This library used to do that, but moved to using Node's native module hooks, so that there's nothing magical going on, and you can still use the `import` expression to import your HMR modules, they just auto-update when saving.

That's orthogonal, and in fact you probably would use TypeScript to translate JSX to JS when using this library. What this does is (a) provide a Node.js module hook to call your transpile function when it encounters TSX/JSX files, and (b) provide a Node.js module that lets you remap imports, including "react/jsx-runtime" if you want a different JSX implementation.

It's English, it just looks like regex. In English, the ? belongs inside the parens in this case.

Just updated the text to be hopefully much clearer.

Oh no, I must have mis-explained it.

The file `site/myfile.js` does exist. All FileTree does is recursively load all files in a dir into memory.

The `useTree` module hook does two things:

* Pulls the file from memory when loading it instead of from disk

* Adds a cache busting query string when resolving it for invalidation

Combined with tree.watch(), this essentially allows you to add a very lightweight but extremely accurate hot module replacement system into Node.js

    const tree = new FileTree('src', import.meta.url)
    registerHooks(useTree(tree))
    tree.watch().on('filesUpdated', () => import(tree.root + '/myfile.js'))
    import(tree.root + '/myfile.js')

Now save src/myfile.js and see it re-executed

Just added some code samples, thanks for the suggestion.

Can't remember, was like 10 years ago. But basically after I finished mastering the last feature I needed to, probably macros.

To clarify, I don't mean that Rich didn't also have extremely good reasons to make Clojure, given he was using Java (and maybe C++) in 2007. They're not the best languages now, but they were so broken back then that they practically caused the language revolution that caused Clojure and Go and Node etc to flourish.

I thought it was self explanatory. It had new idioms I had not yet learned and internalized, so I fully absorbed it. When that was finished, I needed something else to do the same thing with. It's like listening to a song on repeat 10-100 times (depending on the song) when you first hear it. You get everything you can out of it and move on when it's empty.

I guess that confirms my theory.