Rust recently released 1.0, so it’s a perfect time to see what the language has to offer.
To get to know the language, I wrote a multi-threaded web crawler that finds any broken links on a website. In this blog post, I explore the Rust programming experience.
The dependency tooling
The first striking feature of Rust is Cargo. Cargo is a package manager for Rust, and also generates project skeletons to get you started.
Cargo works amazingly well. It’s a stark contrast to programming in C and C++, providing canonical answers to questions like:
- How do I depend on external libraries?
- How do I upgrade those dependencies when new versions are released?
- How do I discover libraries that could help me with my current project?
- How do I work on two projects which depend on different versions of the same library?
- How do I distinguish between direct dependencies I need and indirect dependencies that I happen to use?
If Rust had no other interesting features, the usefulness of Cargo makes Rust very attractive. The closest alternative is biicode, a C/C++ dependencies manager that hasn’t gained much traction. By contrast, the Rust package repository is now growing faster than Haskell’s hackage!
The Rust docs are excellent. They’re cross-referenced, searchable and include direct links to the relevant source code. I’m left wondering why more languages aren’t like this.
Generating docs from source code is hard to get right. I’ve worked with documentation tools that generate elaborate inheritance diagrams, without asking whether it’s the best way to explain your library to a user. At the other extreme, some doc tools require you to write elaborately formatted comments that distract from the code itself. Rustdoc strikes a good balance, using a lightweight markdown syntax.
Virtually every feature in the standard library includes an example in the docs, which is delightful. With one click, you can immediately run the example in the Rust playground!
The Rust community goes to great lengths to be welcoming and inclusive. The code of conduct is applied everywhere, and it shows. For example, here’s a user being asked not to make unconstructive criticisms of Go. This creates a community that I want to be part of.
I was also amazed by how active the community is, especially for such a young language. I made a suggestion for improving a compiler error message and a pull request was merged in under a day! There’s also ample help available with knowledgeable people on IRC and Stack Overflow who are very amenable.
The learning curve
There’s a lot to learn with any new language, but particularly so with Rust. Developers without systems programming experience will be exposed to the distinction between the stack and the heap. There’s also the trait system, which provides object-oriented style encapsulation but without traditional inheritance. It works well, but it’s a novel approach for many developers.
Rust’s biggest conceptual hurdle is the ownership system. This formalises a concept that C/C++ developers already needed to worry about, but takes time to grok. Fortunately, the official Rust book is extremely approachable and broken up into small, digestible chapters. It’s not always easy to make the compiler happy, but the compiler tries hard to make useful suggestions. I was particularly impressed with rustc’s suggestion in this example:
foo.rs:16:24: 16:33 error: attempted to take value of method `get_score` on type `Player` foo.rs:16 let score = player.get_score + 1; ^~~~~~~~~ foo.rs:16:24: 16:33 help: maybe a `()` to call it is missing? If not, try an anonymous function
Rust has a reasonable range of libraries and I had no problems finding what I needed. However, several libraries required me to use a bleeding-edge, nightly build of Rust rather than v1.0. This can be inconvenient because nightly may change an API and you need to downgrade your compiler until the libraries have been updated to match.
Rust compile times aren’t great either (though very competitive with C++). This is an active area of development and it should improve. In the meantime, rest_easy will notify you as soon as the compiler is happy with your code, so you don’t need to wait for code generation.
The end result
Once you’ve finished coding a project, what have you gained? You’re left with something remarkable: robust code (like Haskell) and great performance (like C++). I’ve greatly enjoyed hacking on a Rust project and wouldn’t hesitate to recommend it.
Rust is way past critical mass. The future looks extremely bright.