Rust 0.10
46867cc3

Language FAQ

There aren't many large programs yet. The Rust compiler, 60,000+ lines at the time of writing, is written in Rust. As the oldest body of Rust code it has gone through many iterations of the language, and some parts are nicer to look at than others. It may not be the best code to learn from, but borrowck and resolve were written recently.

A research browser engine called Servo, currently 30,000+ lines across more than a dozen crates, will be exercising a lot of Rust's distinctive type-system and concurrency features, and integrating many native libraries.

Some examples that demonstrate different aspects of the language:

You may also be interested in browsing GitHub's Rust page.

Yes. All development happens in lock-step on all 3 target platforms. Using MinGW, not Cygwin. Note that the windows implementation currently has some limitations: in particular tasks cannot unwind on windows, and all Rust executables require a MinGW installation at runtime.

It is multi-paradigm. Not everything is shoe-horned into a single abstraction. Many things you can do in OO languages you can do in Rust, but not everything, and not always using the same abstraction you're accustomed to.

Data values in the language can only be constructed through a fixed set of initializer forms. Each of those forms requires that its inputs already be initialized. A liveness analysis ensures that local variables are initialized before use.

In short, because too few guarantees could be made about the dynamic environment of the catch block, as well as invariants holding in the unwound heap, to be able to safely resume; we believe that other methods of signalling and logging errors are more appropriate, with tasks playing the role of a "hard" isolation boundary between separate heaps.

Rust provides, instead, three predictable and well-defined options for handling any combination of the three main categories of "catch" logic:

Cleanup through RAII-style destructors is more likely to work than in catch blocks anyways, since it will be better tested (part of the non-error control paths, so executed all the time).

We want to maintain the option to parametrize at runtime. We may make eventually change this limitation, but initially this is how type parameters were implemented.

Doing so would make type inference much more complex, and require the implementation strategy of runtime parametrization.

We don't know if there's an obvious, easy, efficient, stock-textbook way of supporting open or structural disjoint unions. We prefer to stick to language features that have an obvious and well-explored semantics.

There's a lot of debate on this topic; it's easy to find a proponent of default-sync or default-async communication, and there are good reasons for either. Our choice rests on the following arguments:

Similar to the reasoning about default-sync: it wires fewer assumptions into the implementation, that would have to be paid by all use-cases even if they actually require a more complex communication topology.

The str type is UTF-8 because we observe more text in the wild in this encoding -- particularly in network transmissions, which are endian-agnostic -- and we think it's best that the default treatment of I/O not involve having to recode codepoints in each direction.

This does mean that indexed access to a Unicode codepoint inside a str value is an O(n) operation. On the one hand, this is clearly undesirable; on the other hand, this problem is full of trade-offs and we'd like to point a few important qualifications:

In each case there is one or more operator, literal constructor, overloaded use or integration with a built-in control structure that makes us think it would be awkward to phrase the type in terms of more-general type constructors. Same as, say, with numbers! But this is partly an aesthetic call, and we'd be willing to look at a worked-out proposal for eliminating or rephrasing these special cases.

Yes. Calling C code from Rust is simple and exactly as efficient as calling C code from C.

Yes. The Rust code has to be exposed via an extern declaration, which makes it C-ABI compatible. Such a function can be passed to C code as a function pointer or, if given the #[no_mangle] attribute to disable symbol mangling, can be called directly from C code.

For simplicity, we do not plan to do so. Implementing automatic semicolon insertion for Rust would be tricky because the absence of a trailing semicolon means "return a value".

Short answer set the RUST_LOG environment variable to the name of your source file, sans extension.

rustc hello.rs
export RUST_LOG=hello
./hello

Long answer RUST_LOG takes a 'logging spec' that consists of a comma-separated list of paths, where a path consists of the crate name and sequence of module names, each separated by double-colons. For standalone .rs files the crate is implicitly named after the source file, so in the above example we were setting RUST_LOG to the name of the hello crate. Multiple paths can be combined to control the exact logging you want to see. For example, when debugging linking in the compiler you might set RUST_LOG=rustc::metadata::creader,rustc::util::filesearch,rustc::back::rpath

If you aren't sure which paths you need, try setting RUST_LOG to ::help and running your program. This will print a list of paths available for logging. For a full description see the language reference.