A data race that doesn't compile

https://corentin-core.github.io/posts/ruxe-type-level-disjointness/

by stmw • 27 days ago

26 8 27 days ago

corentin-core.github.io

nynx• 27 days ago

I don’t think this needed to be parallelized.

asdf88990• 27 days ago

AI written article. I hate how I can quickly pick this up these days and I hate how I see it everywhere recently.

If you don’t have time to write, I don’t have time to read it.

TonyAlicea10• 27 days ago

I stopped reading at "Here’s the observation that started everything:".

bschwindHN• 27 days ago

No kidding, it drives me crazy. The tone is always so authoritative and dramatic, like every fact is a mic drop moment.

I need to get back to writing some of my own articles I've been meaning to write, as a palate cleanser.

cptroot• 27 days ago

I know this is a small point, but you likely mean "race condition" here. Rust already guarantees safe code to be free of "data races", and this blog post doesn't discuss unsafe in any way.

Two other points that came up while I was reading:

1. I get that this is something you did once for an internal project, but this is what macros and derive_macros are made for.

2. The AI that you used to help write the post left its mark a little too deep. The cliched phrases make an appearance at the start of every section and it really starts to grate by the end.

infinitebit• 27 days ago

“The second condition falls out of the first”. Hi claude!

corentin-core• 15 days ago

You're right about the terminology, my title oversells it. There's actually no concurrent mutation at all: each reducer reads the shared state and returns a fresh slice, and the pieces are stitched back together after. What the compile-time check really guarantees is that every state slice has exactly one reducer. That's what makes running them in parallel safe without a lock.

1. Well no actually, the internal project was in Python, I wanted a playground project inspired by my experience to learn Rust. On macros: I know derive could cut the boilerplate, but I wanted to understand the underlying mechanism first, so I wrote the type-level part by hand. The compile-time guarantee comes from those bounds, not from the macro, so learning that part felt more useful than hiding it. Derive is on my list for later.

2. On the writing: I drafted with Claude (noted at the bottom of the post) and I agree that I didn't edit the tone down enough. Not an excuse, but with a full-time job and two kids I'd rather spend my time on the ideas than the prose. I'll be more careful next time.

ydj• 27 days ago

Surely you can express this in C++ just as you could in rust…

A data race that doesn't compile

Related Stories