Rust

6168 readers

7 users here now

Welcome to the Rust community! This is a place to discuss about the Rust programming language.

Wormhole

[email protected]

Credits

The icon is a modified version of the official rust logo (changing the colors to a gradient and black background)

founded 2 years ago

MODERATORS

[email protected]

-1

When Zig is safer and faster than (unsafe) Rust (zackoverflow.dev)

submitted 1 week ago* (last edited 1 week ago) by erlend_sh to c/[email protected]

6 comments fedilink hide all child comments

There are endless debates online about Rust vs. Zig, this post explores a side of the argument I don't think is mentioned enough.

Intro / TLDR

I was intrigued to learn that the Roc language rewrote their standard library from Rust to Zig. What made Zig the better option?

They wrote that they were using a lot of unsafe Rust and it was getting in their way. They also mentioned that Zig had “more tools for working in a memory-unsafe environment, such as reporting memory leaks in tests”, making the overall process much better.

So is Zig a better alternative to writing unsafe Rust?

I wanted to test this myself and see how hard unsafe Rust would be by building a project that required a substantial amount of unsafe code.

Then I would re-write the project in Zig to see if would be easier/better.

After I finished both versions, I found that the Zig implementation was safer, faster, and easier to write. I’ll share a bit about building both and what I learned.

top 6 comments

sorted by: hot top controversial new old

[–] [email protected] 26 points 1 week ago* (last edited 1 week ago) (2 children)

Title is clickbait. They only talk about unsafe rust, which I can see zip being safer/easier than unsafe rust. But 99.9% of code I write is safe rust - which most people just call rust. Even the author calls out the vast difference to writing safe vs unsafe rust:

Overall, the experience was not great, especially compared to the joy of writing regular safe Rust.

Then I would re-write the project in Zig to see if would be easier/better.

Of course it will be. The second time you write any project it will be easier and faster as you learn a lot form the first time you write something. If zig is always the rewrite it will come off better. Almost all rewrites are better or faster, even if you are moving to a slower language - the language makes a difference to performance and ease of writing. But far more does how you write things and the data structures/algorithms you use.

Overall they seem to want to write as much unsafe as they can and are writing rust like it is C. This is not a good idea and why zig will be better suited. But you can write a VM without large amounts of unsafe if you want to and it can be performant. Unsafe can be used in small parts where performance matters and cannot be done without it (though this is not that common I find).

[–] [email protected] 3 points 1 week ago (1 children)

I don't think the title is that clickbaity, since it's pointing out that there are places where Rust falls short. As with all tools, you should know when to not use them. If you know you will end up with a lot of unsafe code, then Rust may not be a good choice.

The TLDR doesn't say it but the article specifically points out the problem is when trying to use mark-sweep garbage collection in the VM, and talks about the problems that arise when you are forced to use C-like writing to combat the Borrow Checker. Apart from the rewrite in Zig being able to iterate on experiences from the Rust version, I think the points made out were perfectly good arguments about why writing unsafe Rust is far from a good experience. I see that experience reflected in the choice done for the Roc standard library too.

[–] [email protected] 6 points 1 week ago (1 children)

IMO it is clickbaity because it promises to compare rust and zig, but in reality it is just comparing unsafe rust and zig. IMO all it really needed to not be is the word unsafe in the title like they use everywhere else in the article. That is fundamentally my only problem with it. I do agree with your other points on knowing when a tool is good or not to use but I would have been much less likely to click on it if it mentioned unsafe in the title - I already know rusts unsafe is not the best and was expecting some arguments around the advantages of zig over safe rust -ie what most people write, not a small subset of the language.

[–] [email protected] 1 points 1 week ago

I don't think it's fair to consider unsafe Rust such a small subset of the language that it requires calling it out as a separate thing from "Rust" in the title of an article. Unsafe constructs are necessary in the standard library and many crates, whether or not you use it in the code you actually write.

[–] [email protected] 2 points 1 week ago* (last edited 1 week ago) (1 children)

But 99.9% of code I write is safe rust - which most people just call rust.

That's true for your application maybe, but they go on to say how one should consider whether or not their problem is going to fit well within the rules of the rust borrow checker and that needs to be talked about more (vs just assuming Rust is the safest option).

The second time you write any project it will be easier and faster as you learn a lot form the first time you write something. If zig is always the rewrite it will come off better. Almost all rewrites are better or faster, even if you are moving to a slower language - the language makes a difference to performance and ease of writing. But far more does how you write things and the data structures/algorithms you use.

I'm going to agree to disagree with you there. I'd throw this in the category of persistent myth. Yes, ideally, you learn from your first experience and make everything better, but the reality is you often just end up with different mistakes.

Rewrites aren't often done outside of hobbyish projects because they're very expensive, stop new feature development, and you really can end up with something that's worse than what you started with (this is especially true if you've switched languages or frameworks).

Overall they seem to want to write as much unsafe as they can and are writing rust like it is C. This is not a good idea and why zig will be better suited.

They do explain ~~with citations~~ why it makes more sense (i.e. you end up with something more performance) to write their VM outside of the restrictions of the borrow checker.

But you can write a VM without large amounts of unsafe if you want to and it can be performant.

I think the claim is a bit of a stretch off the cuff. Ideally to retort this some rustacian would implement a mark and sweep VM in Rust with maximal use of the borrow checker.

Edit: Looking at their code, it's not all just one big unsafe block either. But it is something that they frequently had to drop down to, to implement this particular garbage collection strategy.

[–] [email protected] 7 points 1 week ago

That’s true for your application maybe, but they go on to say how one should consider whether or not their problem is going to fit well within the rules of the rust borrow checker and that needs to be talked about more (vs just assuming Rust is the safest option).

Yeah, that is why I was calling out their title as click bait. Saying Rust vs Zig will mean most people think of safe rust vs zig. But the article is about unsafe rust vs zig which is a completely different story IMO. If you need lots of unsafe zig might be better - but the title does not say that. Hence IMO it is clickbaity.

Rewrites aren’t often done outside of hobbyish projects because they’re very expensive, stop new feature development, and you really can end up with something that’s worse than what you started with (this is especially true if you’ve switched languages or frameworks).

Ok I will give you that for big projects. But this is on the other side of that. If you write two things in quick succession you will more likely still have the problems encountered in mind and be better able to navigate them the second time than if you are part of a large team that has had a lot of turn over since the project was first written. And you might make different mistakes - but the second round of mistakes is normally less impactful then the first set.

They do explain with citations why it makes more sense (i.e. you end up with something more performance) to write their VM outside of the restrictions of the borrow checker.

Oh yeah, of course this case I can see all that being true. But fundamentally they chose this problem because of that given they said:

I wanted to test this myself and see how hard unsafe Rust would be by building a project that required a substantial amount of unsafe code.

Sounds like they want to compare unsafe rust with zig. They started with that idea which from my experience is not typical of most applications so their findings are not either.

I think the claim is a bit of a stretch off the cuff.

Yeah that one might be.

Though unsafe rust overall is not a large amount of what rust code is, the author does seem to be picking the topic based on the hardest parts of rust and their title talks about all of rust which IMO is an unfair comparison. The story of zig being better than unsafe rust is interesting but only part of the whole zig vs rust debate.