Zig 0.3.0 Released
ziglang.org
ziglang.org
Two minor things:
> zig init-lib and init-exe
Wouldn't these be better off as "zig init lib" and "zig init exe"? I.e. do what git did when it switched from stuff like "git-annex" to "git annex": switch from compound command names to subcommands. Easier mentally to follow and remember.
There's also this:
> all control flow occurs exclusively via keywords.
This will probably never happen cause C looms too large in Zig's world, but I wish "*" and especially the super non-intuitive "&" go the way of the dodo and some short keywords replace them. Perhaps "ptr" and "deref"? This is technically just bikeshedding, but look at Rust threads: besides the obvious #1 difficulty of the borrow checker and ownership, the #2 complaint is that the syntax is weird due to sigils. "Modern" programming languages don't use many sigils, if any: Javascript, Python, Java, C#, etc.
It's not fair since those languages are GCed, but most people will be familiar with them, not with C, these days.
I know that it's doubly unfair since Zig's target is C, so it's expected that new Zig developers know C, but you want to get new developers wherever you can get them, in my opinion :)
More generally using C-style syntax means that C and C++ coders will be at home, while people from other languages will have to familiarize themselves with the concept anyway, sigil or not.
Beyond that, I don't think I've seen any serious study on the acceptable level of sigils to maximize productivity and readability in source code, as far as I can tell it's really just a matter of opinion. Personally I'd take C syntax over Python every day of the month, and I've written quite a bit of Python. My only sigil-related problem with C is that '->' should really just be '.' IMO.
Beware Stroustrup's Rule[0] :).
[0] https://thefeedbackloop.xyz/stroustrups-rule-and-layering-ov...
On a more serious note, I think that this word form is concise enough and might emphasize code smells in a more modern language, where stuff like * *a should be super rare, IMHO.
I played around with it some months ago. It was an adventure so to speak. Even discovered bugs in LLVM itself. Andy was very responsive to bug requests, so I got a demo working pretty quickly.
I'm hoping to do another pass soon, perhaps contribute a bit to the compiler.
The demo is here: https://github.com/skyfex/zig-nrf-demo
Probably doesn't work anymore.. but that's the way it goes with a developing language :) That's part of the fun
> Zig intends to be so practical that people find themselves using it even if they dislike it.
feels awfully condescending to my ears personally, in an I'm-smarter-and-know-better-than-you kind of way, even if you didn't intend it like that.
Just my unsolicited $0.02, but I'd recommend a more objective and non-loaded description something along the lines of "Zig's primary motivation is practicality, and therefore avoids issues with C such as <x, y, z>".
I'm sure this an excellent language, but it sounds like it's written by a megalomaniac.
I'm sure it's not intended that way, but I would advise the author to tone it down a bit so people don't take it the wrong way.
Previous HackerNews discussion on Zig: https://news.ycombinator.com/item?id=17184407
Also, perhaps not as objective as you'd like but still good resources:
Whirlwind tour of the basic thinking behind Zig: https://andrewkelley.me/post/intro-to-zig.html
A ~50 minute talk on Zig, by its creator: https://www.youtube.com/watch?v=Z4oYSByyRak
The long of it: The competition right now in this space as I see it is: C, Rust, C++. Even C++ doesn't fully fit because you have to turn off language features like exceptions to have the guarantees you need for systems programming. So in turn.
C hasn't really changed in 40 years so all it's warts are still intact and has specific pain points that will never be addressed.
C++ conversely has added every language feature plus the kitchen sink over the past 30 years. People say: "Just don't use the features you don't want", but that is unrealistic when you have to interoperate with third party libraries etc. If you have to drop a large portion of the C++ ecosystem to use the subset of the language you like then you've lost a major benefit of choosing C++ in the first place.
Rust, while being more thoughtful about it than C++, has provided a large set of features to work with. It is a much bigger language than C. It's also taken a very opinionated stance on safety with the borrow checker which causes a large learning curve and a lot of mental overhead.
Finally Zig is an attempt to be more safe than C where it can without being as overbearing as Rust. Add just the necessary features that a huge boost to quality of life while not flooding the programmers mental model of the language. I think of it as kind of a C+.
It imposes design constraints that in the end work out for the better. This is why Rust users are so fanatical about it.
I'd also add that there are (small) warts on Rust's current implementation of borrow checking which everyone agrees are minor annoyances, even to experienced users:
https://github.com/rust-lang/rust/issues/43234
Judging by that issue thread, some of these issues are about to be fixed.
I feel the jury is still out on this. It does impose design constraints that remove a certain subclass of errors, whether this correlates to actually causing better designed programs in the sense of readability, maintainability, etc. I'm unsure. I understand the desire for this to be true and I feel that some of the fanaticism around the language is the belief that this is true. Developers like to have the 'one true way' to solve problems. Go's popularity I feel is in part because of this as well but achieved it by having a bare bones language with less options instead of a restrictive compiler. If you take the 'restrictive compiler' argument to its logical conclusion for a program that accepts input x and produces output y there would only be one valid way to code this program. This would be good in the sense that presumably the compiler has proven that your program works and is correct. On the other hand though I have no inclination to believe that this program would be the most understandable, modifiable way to get from x -> y.
What are some of those?
Looked it up, found these links, for anyone interested:
https://blog.regehr.org/archives/490
https://stackoverflow.com/questions/11276259/are-data-races-...
But, it'll still be orders of magnitude larger than the Zig ecosystem is likely to be anytime in the next couple of decades.
I'm not saying one shouldn't use Zig or that one should use C++...I've never been able to grok C++, as it is far too large for the small amount of time I have to spend on learning about systems languages (I can barely read it, much less write it effectively). I just think it's weird to argue for Zig with a point about ecosystem size. Zig will likely never look good on that metric, compared to C or C++.
Personally I cheer for Zig and think that it has the biggest potential to really become a C replacement for me. Other then that I hope that Myrddin [0] will see stable development till 1.0 and beyond. Myrddin has my vote, because it uses QBE [1] compiler backend instead of LLVM. Now that I think about it it would be cool to see Zig on QBE. I just don't think that the current LLVM monoculture is healthy in a long run.
Maybe I'm weird, but I would like to imagine a language like Zig compiled also to DEX for Android and CIL for Windows. Then it could have mostly free platform-native calls. So one could really write truly first-class cross-platform apps.
from https://c9x.me/compile/:
> Text input only. A C library interface is work in progress.
this is a bit of a non-starter
> Some ongoing work on the Myrddin compiler aims to use QBE as backend.
sounds like myrddin does not use QBE as a backend.
----
It is unlikely that Zig will ever use QBE as a backend, optional or otherwise. However, it is quite possible that in some distant future Zig has its own backend that avoids the LLVM dependency entirely. This would be in pursuit of fast compile times.
It's great to hear that a home grown backend is something even considered for future. I know that it's a long way there, but that's life.
Thank you for your efforts with Zig. From all the new systemish languages I find Zig with it's current momentum the most probable to become the C replacement for me.
This is a serious project that I'm working on. The 'engine' uses the Vulkan graphics API, cache efficient design, scales to the number of cores, game logic is multithreaded, AVX is used, and so on.
I only mention this to hopefully get people interested in Zig. I'm not ready to share screenshots yet and I'm certainly not going to be advertising my game here.
I've been trying out all of these languages but haven't decided which one to dive into completely.
I'd also be curious why the OP decided not to use Rust.
1) What's your workflow like in terms of tooling? Coming from C# and C++ people are typically used to great IDE support, debugging and profiling tools, etc. Zig doesn't have any of that. How do you work around it and do you find the experience reasonably productive?
2) How do you deal with the language still rapidly evolving? Do you run the latest nightlies or freeze a particular compiler version and update periodically?
3) What are the most important things Zig gets you over the traditional gamedev approach of using a restricted subset of C++ (in some cases literally just C with function/operator overloading for basic math and a few other convenience features)?
2) I git pull every few days to run on the latest compiler. The bugs that I've hit have been minor inconveniences, nothing serious. I think Andrew would agree with you though that Zig is probably too immature for a serious project at this point. The transition to the self-hosted compiler has yet to take place, that compiler is in its infancy.
3) I'm afraid that my verbal inability will never do Zig justice. The built in slices and optionals are really nice. Pointers are non-null by default (but can be undefined). Stack traces on crashes. Much better casting in my opinion (once I got used to it). Error handling that's similar to Herb's C++ deterministic exception proposal. No headers. Not having to decide between fast compile times and comfortable printing. Defer statements instead of destructors or nothing work really well in my opinion. Built in testing (maybe not something you use much in gamedev). +% means add with overflow, + means overflow is undefined. Everything feels nicer, that's as good as I can put it.
edit: You can Ctrl+F "pub fn" to browse all public functions. Great when browsing Github. It's easy to grep for the definition of anything. Overflow and other undefined behavior is detected in debug mode, this saves me so much time. Guaranteed inline is nice. No more make/cmake and manually listing which files to compile.
For SIMD I'm using ispc which is a shader-like language/compiler from Intel. You write your code from the point of view of a single SIMD lane and the compiler takes care of the rest. Intel kind of screwed the project over with their stupid naming and terrible website but it's a very cool project. I take the resulting object file and link it with the rest of the project. This is very different from trusting a compiler to vectorize your code, with a shading language you know it's going to vectorize. What's absolutely brilliant is when you start to use divergent control flow and see how well it handles it.