Comments on it:
news.ycombinator.com/item?id=4318...

blog.mozilla.org/en/products/...

I try to reproduce this by using all available modifier keys and tab, but it just works for me.
The next version will have the entire area clickable, which means that if this happens again, you can just click it and set a new modifier key combo.
Still, it's interesting why it's happening.

Thank you! I had a lot of fun and am looking forward to continuing to use and contribute to Hex.

This worked for a small file, but as soon as I increased above a megabyte, the C implementation would output incorrect data. Turns out, I forgot that malloc doesn't guarantee 0 initialized memory, which I relied on. Easy fix, and now it's ready for advent of cancer!

With the compiler running, I could write the challenge itself. Generate the input data in JS, write them to a file in the wasmer runtime, call clang on the user code, run the user code, read back the output file and finally compare to the reference implementation.

A little bit of js and html later and the first issue became obvious: I need Cross-Origin policies, but the server I use for running locally doesn't offer custom headers. Easy fix: register a webworker and add the headers there.

A few of the challenges are going to be optimization problems, so we need an easy solution for running compiled code. I thought about setting up some kind of runner, until I remembered wasm and looking around, there is an easy to use c to wasm compiler for wasm using wasmer. wasmer.io/clang/clang

Fixing that would require a rewrite of the ast to remember what whitespace the user put in, then making smart decisions about what to keep and what whitespace to insert.
That of course also means baking opinions into the formatter, but it's my language, so that par is easy.

For example, it always puts a newline after the "{" in a class definition, which results in the empty class being two lines long.
It also completely ignores how long a line is going to be because of things like arguments in a function declaration, which could be spread out on multiple lines.

For anyone interested, you can follow along at
github.com/luksab/rlox
Or read the work in progress blog post on my website emmabyte.de/things-i-lea...

Not that it's important for now, but the goal I have in mind is something along the lines of "the goals of Golang with the good design of Rust".

Once I'm done redoing the structure, I want to "make it mine", meaning change the syntax to me more like what I like (Rust) and the semantics to allow for the kind of code I have in mind.

It really clicked when I started refactoring yesterday. About half the codebase is now prettied up, but I hope to continue throughout the weekend.

I was not around before Ai was done with machine learning, but it sounds fun

Angy metal birds feeding. Nature is healing