Working with strings in Rust (2020)
fasterthanli.me
fasterthanli.me
> Okay, so, argv is an array of addresses, and at those addresses, there is.. string data. Something like that:
> It looks like every argument is terminated by the value 0. Indeed, C has null-terminated strings.
> If I'm, uh, reading this correctly, "é" is not a char, it's actually two chars in a trenchcoat. That seems... strange.
The inclusion of all the "naive test programs" serves no purpose but to inflate the article's length to ridiculous levels.
Lots of complete working examples let me test my working mental model I develop while learning too.
It may or may not be relevant that I've been reading TAOCP assembly programs recently so may be thinking more in terms of pointers.
(I think the intended audience is something like "people who want to read about things they already sort-of-know, set out as a nice story to help everything fit into place". The apparent audience of "people who are comfortable reading C but don't know that C strings are null-terminated" is presumably a rather small one.)
However, I think the way the language has implemented `async` is going to cause a riff in the usage. Because they didn't standardize the executors interface, you're essentially forced to using a single runtime (Tokio) since library developers would have to code for each runtime. Also, because of the async keyword, library developers would also have to make a different sync/async of each function they want to declare.
These are not small issues and will only continue to get worse as people are forced to choose one route, leaving others to use a different language.
The one saving grace with Rust is if everyone decides to say "screw async" and just builds synchronous APIs, then we use something like [May](https://github.com/Xudong-Huang/may) for green threading.
> Kidding, I don't have a plan.
:|
This is a program with async in general, no? Go avoids this by just having everything be async, but that seems unsuitable for Rust since you won't have control over when your code executes and what thread it executes on, not a problem for applications but it would seem to rule out things like kernel modules and firmwares, as well as require every type to be Send + Sync (which ultimately means the language must be garbage collected).
In Python for example I've written async and sync versions of a library, and some magic testing that makes sure both implementations return the same results.
> These are not small issues and will only continue to get worse as people are forced to choose one route, leaving others to use a different language.
Honestly this doesn't seem like a problem to me. I think there's room for many languages in this world. Rust is a wonderful language but it isn't a "one true language." People should choose a different language if it suits them better. That's no more an indictment of Rust than it's an indictment of Python when I embed Rust in a web app - that's not a place where Python shines.
No. If you are an async program calling a sync library, you can use `tokio::task::spawn_blocking`, and Tokio will run your sync function on a worker thread that's allowed to block https://docs.rs/tokio/latest/tokio/task/fn.spawn_blocking.ht...
The `reqwest` library offers a sync API by just wrapping its own async API: https://docs.rs/reqwest/0.11.12/src/reqwest/blocking/client....
For libraries that don't do this convenience wrapping, you can just create an async runtime like `tokio::Runtime` and use a function like `block_on`: https://docs.rs/tokio/latest/tokio/runtime/struct.Runtime.ht...
Since all Rust programs have sync entrypoints, there is always some way to call async functions from sync functions. It's not like C# or Go where the async runtime is a singleton built deep into the language. You always have control of it.
In practice, if I'm doing anything with networking, I always end up wanting async, so I prefer to have the runtime and use spawn_blocking for sync APIs.
Not everyone wants to pull in an async runtime just to run make a call to an API.
My whole point is async puts a huge burden on library authors that will force users to diverge in their approaches and that will ultimately hurt the community
To me C++ pretty well follows the "make simple things easy/safe, and hard things possible" philosophy. I'm glad this it still does support low level stuff, but the use cases for that are minimal and I do wish there was a compiler flag to reject full C backwards compatibility and only support a modern safe subset, so that projects could protect themselves from junior developers shooting themself in the foot.
It's a shame that C++ doesn't provide mt-safe versions of STL types in the standard library, although trivial to wrap them yourself (1 line of code per method, using std::lock_guard).
Are all Rust data types mt-safe? Does unsafe mode provide faster unprotected versions if you need those?
Yes, but not in the sense you probably think, given your second sentence:
> Does unsafe mode provide faster unprotected versions if you need those?
Some data types can be safely shared between threads, and some cannot. Rust checks at compile time if you try and use a non-thread safe data structure from multiple threads, and if you do, will give you an error. So in that sense, all of them are safe, yes.
You don’t use unsafe to get access to non-thread safe data structures, you may have both kinds, and the compiler checks you use them correctly.
Is there any provision for building your own thread-safe types (e.g. a structure composed of other types) out of non-thread-safe types and mutexes, and if so how does that work in terms of compile-time errors ?
To my knowledge, no. If you want to push to a vector from multiple threads, you "wrap" the vector in a `Mutex`. The difference between C++'s std::mutex or std::lock_guard and Rust's Mutex, is that the compiler refuses to let you touch the data protected by the mutex unless you have a lock acquired on the mutex.
> Is there any provision ...
Yes!
1. In Rust, types can implement "traits".
2. There's two traits that control thread safety: `Send`, and `Sync` Basically, any type that "implements the trait" `Send` is safe to send across threads. And a reference of any type that implements `Sync` is safe to to send across threads.
3. `Send` and `Sync` are "auto-traits", which means that if you make a data structure out of primitives that all implement `Send`, your data structure will also implement `Send`, same for `Sync`.
4. There are a bunch of thread-safety primitives that you can use (like `Arc` (atomic reference counters), `Mutex`, ..etc) to build thread safe data structures.
> how does that work in terms of compile-time errors?
The compiler will not let send a type across threads if it doesn't implement `Send`, and it won't let you send a reference to a type if the type doesn't implement `Sync`!
That way, if you avoid using the `unsafe` keyword, and the compiler agrees to compile your code, you can be sure you won't have data races!
In Rust, there are traits (you might know them as interfaces or protocols from other languages) called Send and Sync which tell the language whether or not something can be send to a different thread, or shared between threads.
Vec<T> is Send but not Sync - you could pass ownership of it to a different thread, surrendering your access to it in the process, but two threads couldn't share access to it. Almost everything is Send, unless it contains references to some thread local state.
Mutex<Vec<T>> is Sync - you can share it between threads. Basically if you take the lock, you'll get back a smart pointer to your Vec<T>.
These traits are generally implemented automatically; you can implement them on a type yourself, using unsafe, but you'd only do that if you were writing your own synchronization primitives or thread safe data structures.
So; the compiler is able to infer which types are and are not thread safe, and what flavor of thread safety it has. It's then able to use that to check for thread safety violations at compile time.
There's more to it than that, in particular it's possible to share a read-only reference to a Vec<T> between threads (with a compile time guarantee that no one has a mutable reference to it), but I'd refer you to the book[1] or to other articles on HN of you wanted the specifics.
Personally, I don't think it's worth all that just to avoid a garbage collector, but it's not like there are that many decent compile languages with GC currently on the hype train.
I had C/C++ in college and spent most of my career in front-end.
In comparison to the Cs Rust at least tries to get out your way. Stuff generally compiles and when it doesn't, it tells you what might be wrong using human language.
I mean, it's the first time I've seen a compiler error start with "perhaps..." to tell you what steps could be taken to fix it.
LLVM-based compilers such as the rust compiler and Clang (C++) tend to be way better than g++ in this regard, although Clang competition has forced g++ to be a little better in recent years.
What does this mean?
Is that an accurate summary of your stance?
If you want me to go there, I will say that indeed, saying "just pay attention" is not an answer - "find ways to not let yourself make those mistakes" is already a more valuable piece of advice.
I will also say that one should try this style (that is, program stuff in a type-less language like assembly) rather than dismiss it as "old" or "obsolete". Lack of safety will never be obsolete, so learning how to deal with it is a valuable skill.
Yes, you can always avoid making mistakes, but you can't make others avoid the same mistakes. It just never scales.
The software industry on the other hand is scaling heavily. Which is why what worked in the 80s isn't good enough anymore.
OTOH I experience quite the opposite everyday in the software industry, so I understand one could reject that position as too idealistic.
This makes perfect sense and I do agree. But I also wonder what you would say to someone writing, say, an operating system or a browser? Projects that are inherently large and where the stakes of errors are high.
I do think you need great safety and abstractions to raise your complexity ceiling to the level of these projects, if that's what you've set out to do.
Browser are the typical case of a project/product that became bloated over time; they began as relatively simple document downloader and viewer, but now they are also application downloader and execution environments. This causes all kind of problems for users and implementors. The answer is somewhat eye-rolling: we should go back to Flashplayer times (with something better than FlashPlayer though). Let browsers handle HTML/CSS, and delegate everything else (PDF, big media files,...) to dedicated applications chosen by the end user. One can still do that, indeed, except it has become more and more difficult - even for simple text-and-images documents - because websites are abusing Javascript capabilities.
"we have to handle complexity better" is the common way of thinking, and I think it is wrong as a first reaction. Trying to lower the complexity first is better.
Is your position that no project needs to be large if we employ the right strategy, or that we should look at our complexity ceiling and say, "okay, that's how large a project we can take on?" Or something else?
Thank you for mentioning this book, I look forward to reading it.
"Rust is safer than C" - meh, so what? So is C++ ...
This isn't an article about why you should use Rust, it's about strings in Rust.
Well, the article literally has more C code in it than Rust, so it sure seems to be about comparing Rust to C.
Why the author thought it would be useful to compare Rust's strings to C, a language that literally doesn't even have a string type is beyond me. Kind of like saying "Look how fast I can run! Even faster than a walrus!"
C++ programmers might do OK to think of &str as like a C++ 17 string_view, except it's promised to always be UTF-8 text, and because it was provided from the outset in the language this type is always used where appropriate, so the string literal "Test" is a &'static str, and all the APIs you use work with &str and never some archaic pointer type.
Rust's String is a bit like C++ std::string, though deliberately lacking the Small String Optimisation but again always UTF-8, and String implements the Trait Deref<Target=str> which means most read-only methods are defined on &str but work fine on String anyway.
[edited: A handful of read-only methods make sense specifically for String, like asking how much space is allocated for the String, so those are in fact methods on String]