Origins (2006–2012)

Rust grew out of a personal project begun in 2006 by Mozilla Research employee Graydon Hoare.^[20] Mozilla began sponsoring the project in 2009 as a part of the ongoing development of an experimental browser engine called Servo,^[21] which was officially announced by Mozilla in 2010.^[22][23] During the same year, work shifted from the initial compiler written in OCaml to a self-hosting compiler based on LLVM written in Rust. The new Rust compiler successfully compiled itself in 2011.^[21]

Evolution (2012–2019)

Rust's type system underwent significant changes between versions 0.2, 0.3, and 0.4. In version 0.2, which was released in March 2012, classes were introduced for the first time.^[24] Four months later, version 0.3 added destructors and polymorphism, through the use of interfaces.^[25] In October 2012, version 0.4 was released, which added traits as a means of inheritance. Interfaces were combined with traits and removed as a separate feature; and classes were replaced by a combination of implementations and structured types.^[26]

Through early 2010s, memory management through the ownership system was gradually consolidated to prevent memory bugs. By 2013, Rust's garbage collector was removed, with the ownership rules in place.^[20]

In January 2014, the editor-in-chief of Dr. Dobb's Journal, Andrew Binstock, commented on Rust's chances of becoming a competitor to C++, along with D, Go, and Nim (then Nimrod). According to Binstock, while Rust was "widely viewed as a remarkably elegant language", adoption slowed because it radically changed from version to version.^[27] The first stable release, Rust 1.0, was announced on May 15, 2015.^[28][29]

The development of the Servo browser engine continued alongside Rust's own growth. In September 2017, Firefox 57 was released as the first version that incorporated components from Servo, in a project named "Firefox Quantum".^[30]

Mozilla layoffs and Rust Foundation (2020–present)

In August 2020, Mozilla laid off 250 of its 1,000 employees worldwide, as part of a corporate restructuring caused by the COVID-19 pandemic.^[31][32] The team behind Servo was disbanded. The event raised concerns about the future of Rust, as some members of the team were active contributors to Rust.^[33] In the following week, the Rust Core Team acknowledged the severe impact of the layoffs and announced that plans for a Rust foundation were underway. The first goal of the foundation would be to take ownership of all trademarks and domain names, and take financial responsibility for their costs.^[34]

On February 8, 2021, the formation of the Rust Foundation was announced by its five founding companies (AWS, Huawei, Google, Microsoft, and Mozilla).^[35][36] In a blog post published on April 6, 2021, Google announced support for Rust within the Android Open Source Project as an alternative to C/C++.^[37]

On November 22, 2021, the Moderation Team, which was responsible for enforcing community standards and the Code of Conduct, announced their resignation "in protest of the Core Team placing themselves unaccountable to anyone but themselves".^[38] In May 2022, the Rust Core Team, other lead programmers, and certain members of the Rust Foundation board implemented governance reforms in response to the incident.^[39]

The Rust Foundation posted a draft for a new trademark policy on April 6, 2023, revising its rules on how the Rust logo and name can be used, which resulted in negative reactions from Rust users and contributors.^[40]

Hello World program

Below is a "Hello, World!" program in Rust. The fn keyword denotes a function, and the println! macro prints the message to standard output.^[44] Statements in Rust are separated by semicolons.

fn main() {
    println!("Hello, World!");
}

Keywords and control flow

In Rust, blocks of code are delimited by curly brackets, and control flow is implemented by keywords such as if, else, while, and for.^[45] Pattern matching can be done using the match keyword.^[46] In the examples below, explanations are given in comments, which start with //.^[47]

fn main() {
    // Defining a mutable variable with 'let mut'
    // Using the macro vec! to create a vector
    let mut values = vec![1, 2, 3, 4];

    for value in &values {
        println!("value = {}", value);
    }

    if values.len() > 5 {
        println!("List is longer than five items");
    }

    // Pattern matching
    match values.len() {
        0 => println!("Empty"),
        1 => println!("One value"),
        // pattern matching can use ranges of integers
        2..=10 => println!("Between two and ten values"),
        11 => println!("Eleven values"),
        // A `_` pattern is called a "wildcard", it matches any value
        _ => println!("Many values"),
    };

    // while loop with predicate and pattern matching using let
    while let Some(value) = values.pop() {
        println!("value = {value}"); // using curly brackets to format a local variable
    }
}

Expression blocks

Rust is expression-oriented, with nearly every part of a function body being an expression, including control-flow operators.^[48] The ordinary if expression is used instead of C's ternary conditional. With returns being implicit, a function does not need to end with a return expression; if the semicolon is omitted, the value of the last expression in the function is used as the return value,^[49] as seen in the following recursive implementation of the factorial function:

fn factorial(i: u64) -> u64 {
    if i == 0 {
        1
    } else {
        i * factorial(i - 1)
    }
}

The following iterative implementation uses the ..= operator to create an inclusive range:

fn factorial(i: u64) -> u64 {
    (2..=i).product()
}

|<parameter-name>: <type>| -> <return-type> { <body> };

let f = |x: i32| -> i32 { x * 2 };

let f = |x| { x * 2 };

let f = |x| x * 2;

let f = || println!("Hello, world!");

// A function which takes a function pointer as an argument and calls it with
// the value `5`.
fn apply(f: fn(i32) -> i32) -> i32 {
    // No semicolon, to indicate an implicit return
    f(5)
}

fn main() {
    // Defining the closure
    let f = |x| x * 2;

    println!("{}", apply(f));  // 10
    println!("{}", f(5));      // 10
}

// A function that takes a value of type F (which is defined as
// a generic type that implements the `Fn` trait, e.g. a closure)
// and calls it with the value `5`.
fn apply_by_ref<F>(f: F) -> i32
    where F: Fn(i32) -> i32
{
    f(5)
}

fn main() {
    let f = |x| {
        println!("I got the value: {}", x);
        x * 2
    };
    
    // Applies the function before printing its return value
    println!("5 * 2 = {}", apply_by_ref(f));
}

// ~~ Program output ~~
// I got the value: 5
// 5 * 2 = 10

fn apply_by_ref(f: impl Fn(i32) -> i32) -> i32 {
    f(5)
}

Types

Rust is strongly typed and statically typed. The types of all variables must be known at compilation time; assigning a value of a particular type to a differently typed variable causes a compilation error. Variables are declared with the keyword let, and type inference is used to determine their type.^[53] Variables assigned multiple times must be marked with the keyword mut (short for mutable).^[54]

The default integer type is i32, and the default floating point type is f64. If the type of a literal number is not explicitly provided, either it is inferred from the context or the default type is used.^[55]

let boxed: Box<u8> = Box::new(5);
let val: u8 = *boxed;

let five = Rc::new(5);
let also_five = five.clone();

let foo = Arc::new(vec![1.0, 2.0]);
let a = foo.clone(); // a can be sent to another thread

let c = Cell::new(5);
c.set(10);

let mutex = Mutex::new(0_u32);
let _guard = mutex.lock();

let lock = RwLock::new(5);
let r1 = lock.read().unwrap();

 let (lock, cvar) = (Mutex::new(true), Condvar::new());
// As long as the value inside the `Mutex<bool>` is `true`, we wait.
let _guard = cvar.wait_while(lock.lock().unwrap(), |pending| { *pending }).unwrap();

Duration::from_millis(1) // 1ms

let mut player_stats = HashMap::new();
player_stats.insert("damage", 1);
player_stats.entry("health").or_insert(100);

let mut solar_distance = BTreeMap::from([
    ("Mercury", 0.4),
    ("Venus", 0.7),
]);
solar_distance.entry("Earth").or_insert(1.0);

fn main() {
    let name1: Option<&str> = None;
    // In this case, nothing will be printed out
    if let Some(name) = name1 {
        println!("{name}");
    }

    let name2: Option<&str> = Some("Matthew");
    // In this case, the word "Matthew" will be printed out
    if let Some(name) = name2 {
        println!("{name}");
    }
}

let x = 1000;
println!("1000 as a u16 is: {}", x as u16);

Ownership and lifetimes

Rust's ownership system consists of rules that ensure memory safety without using a garbage collector. At run time, each value must be attached to a variable called the owner of that value, and every value must have exactly one owner.^[85] Values are moved between different owners through assignment or passing a value as a function parameter. Values can also be borrowed, meaning they are temporarily passed to a different function before being returned to the owner.^[86] With these rules, Rust can prevent the creation and use of dangling pointers:^[86][87]

fn print_string(s: String) {
    println!("{}", s);
}

fn main() {
    let s = String::from("Hello, World");
    print_string(s); // s consumed by print_string
    // s has been moved, so cannot be used any more
    // another print_string(s); would result in a compile error
}

Because of these ownership rules, Rust types are known as linear or affine types, meaning each value can be used exactly once. This enforces a form of software fault isolation as the owner of a value is solely responsible for its correctness and deallocation.^[88]

Lifetimes are usually an implicit part of all reference types in Rust. Each lifetime encompasses a set of locations in the code for which a variable is valid. For example, a reference to a local variable has a lifetime corresponding to the block it is defined in:^[89]

fn main() {
    let r = 9;                              // ------------------+- Lifetime 'a
                                            //                   |
    {                                       //                   |
        let x = 5;                          // -+-- Lifetime 'b  |
        println!("r: {}, x: {}", r, x);     //  |                |
    }                                       //  |                |
                                            //                   |
    println!("r: {}", r);                   //                   |
}                                           // ------------------+

The borrow checker in the Rust compiler uses lifetimes to ensure that the values a reference points to remain valid. It also ensures that a mutable reference exists only if no immutable references exist at the same time.^[90][91] In the example above, the subject of the reference (variable x) has a shorter lifetime than the lifetime of the reference itself (variable r), therefore the borrow checker errors, preventing x from being used from outside its scope.^[92] Rust's memory and ownership system was influenced by region-based memory management in languages such as Cyclone and ML Kit.^[5]

Rust defines the relationship between the lifetimes of the objects created and used by functions, using lifetime parameters, as a signature feature.^[93]

When a stack or temporary variable goes out of scope, it is dropped by running its destructor. The destructor may be programmatically defined through the drop function. This technique enforces the so-called resource acquisition is initialization (RAII) design pattern, in which resources, such as file descriptors or network sockets, are tied to the lifetime of an object: when the object is dropped, the resource is closed.^[94][95]

The example below parses some configuration options from a string and creates a struct containing the options. The struct only contains references to the data; so, for the struct to remain valid, the data referred to by the struct must be valid as well. The function signature for parse_config specifies this relationship explicitly. In this example, the explicit lifetimes are unnecessary in newer Rust versions, due to lifetime elision, which is an algorithm that automatically assigns lifetimes to functions if they are trivial.^[96]

use std::collections::HashMap;

// This struct has one lifetime parameter, 'src. The name is only used within the struct's definition.
#[derive(Debug)]
struct Config<'src> {
    hostname: &'src str,
    username: &'src str,
}

// This function also has a lifetime parameter, 'cfg. 'cfg is attached to the "config" parameter, which
// establishes that the data in "config" lives at least as long as the 'cfg lifetime.
// The returned struct also uses 'cfg for its lifetime, so it can live at most as long as 'cfg.
fn parse_config<'cfg>(config: &'cfg str) -> Config<'cfg> {
    let key_values: HashMap<_, _> = config
        .lines()
        .filter(|line| !line.starts_with('#'))
        .filter_map(|line| line.split_once('='))
        .map(|(key, value)| (key.trim(), value.trim()))
        .collect();
    Config {
        hostname: key_values["hostname"],
        username: key_values["username"],
    }
}

fn main() {
    let config = parse_config(
        r#"hostname = foobar
username=barfoo"#,
    );
    println!("Parsed config: {:#?}", config);
}

Memory safety

Rust is designed to be memory safe. It does not permit null pointers, dangling pointers, or data races.^[97][98][99] Data values can be initialized only through a fixed set of forms, all of which require their inputs to be already initialized.^[100]

Unsafe code can subvert some of these restrictions, using the unsafe keyword.^[78] Unsafe code may also be used for low-level functionality, such as volatile memory access, architecture-specific intrinsics, type punning, and inline assembly.^[101]

Memory management

Rust does not use automated garbage collection. Memory and other resources are managed through the "resource acquisition is initialization" convention,^[102] with optional reference counting. Rust provides deterministic management of resources, with very low overhead.^[103] Values are allocated on the stack by default, and all dynamic allocations must be explicit.^[104]

The built-in reference types using the & symbol do not involve run-time reference counting. The safety and validity of the underlying pointers is verified at compile time, preventing dangling pointers and other forms of undefined behavior.^[105] Rust's type system separates shared, immutable references of the form &T from unique, mutable references of the form &mut T. A mutable reference can be coerced to an immutable reference, but not vice versa.^[106]

Polymorphism

use std::ops::Add;

// sum is a generic function with one type parameter, T
fn sum<T>(num1: T, num2: T) -> T
where  
    T: Add<Output = T>,  // T must implement the Add trait where addition returns another T
{
    num1 + num2  // num1 + num2 is syntactic sugar for num1.add(num2) provided by the Add trait
}

fn main() {
    let result1 = sum(10, 20);
    println!("Sum is: {}", result1); // Sum is: 30

    let result2 = sum(10.23, 20.45);
    println!("Sum is: {}", result2); // Sum is: 30.68
}

use std::fmt::Display;

let v: Vec<Box<dyn Display>> = vec![
    Box::new(3),
    Box::new(5.0),
    Box::new("hi"),
];

for x in v {
    println!("{x}");
}

Iterators

For loops in Rust work in a functional style as operations over an iterator type. For example, in the loop

for x in 0..100 {
   f(x);
}

0..100 is a value of type Range which implements the Iterator trait; the code applies the function f to each element returned by the iterator. Iterators can be combined with functions over iterators like map, filter, and sum. For example, the following adds up all numbers between 1 and 100 that are multiples of 3:

(1..=100).filter(|&x| x % 3 == 0).sum()

Macros

It is possible to extend the Rust language using macros.

Procedural macros

Procedural macros are Rust functions that run and modify the compiler's input token stream, before any other components are compiled. They are generally more flexible than declarative macros, but are more difficult to maintain due to their complexity.^[119][120]

Procedural macros come in three flavors:

• Function-like macros custom!(...)
• Derive macros #[derive(CustomDerive)]
• Attribute macros #[custom_attribute]

The println! macro is an example of a function-like macro. Theserde_derive macro^[121] provides a commonly used library for generating code for reading and writing data in many formats, such as JSON. Attribute macros are commonly used for language bindings, such as the extendr library for Rust bindings to R.^[122]

The following code shows the use of the Serialize, Deserialize, and Debug-derived procedural macros to implement JSON reading and writing, as well as the ability to format a structure for debugging.

A UML diagram depicting a Rust struct named Point.

use serde_json::{Serialize, Deserialize};

#[derive(Serialize, Deserialize, Debug)]
struct Point {
    x: i32,
    y: i32,
}

fn main() {
    let point = Point { x: 1, y: 2 };

    let serialized = serde_json::to_string(&point).unwrap();
    println!("serialized = {}", serialized);

    let deserialized: Point = serde_json::from_str(&serialized).unwrap();
    println!("deserialized = {:?}", deserialized);
}

macro_rules! calculate {
    // The pattern for a single `eval`
    (eval $e:expr) => {{
        {
            let val: usize = $e; // Force types to be integers
            println!("{} = {}", stringify!{$e}, val);
        }
    }};

    // Decompose multiple `eval`s recursively
    (eval $e:expr, $(eval $es:expr),+) => {{
        calculate! { eval $e }
        calculate! { $(eval $es),+ }
    }};
}

fn main() {
    calculate! { // Look ma! Variadic `calculate!`!
        eval 1 + 2,
        eval 3 + 4,
        eval (2 * 3) + 1
    }
}

Interface with C and C++

Rust has a foreign function interface (FFI) that can be used both to call code written in languages such as C from Rust and to call Rust code from those languages. As of 2023, a external library called CXX exists for calling to or from C++.^[125] Rust and C differ in how they lay out structs in memory, so Rust structs may be given a #[repr(C)] attribute, forcing the same layout as the equivalent C struct.^[126]

Compiler

The Rust compiler is named rustc. Internally, rustc is a frontend to various backends that are used for further device-specific and platform-specific binary code files (e.g. ELF or WASM binary) generation (compilation). The most mature and most used backend in rustc is the LLVM intermediate representation (bytecode) compiler,^[128] but a Cranelift-based backend and a GCC-based backend are also available in nightly builds of the compiler.^[129]

Standard library

The Rust standard library defines and implements many widely used custom data types, including core data structures such as Vec, Option, and HashMap, as well as smart pointer types. Rust also provides a way to exclude most of the standard library using the attribute #![no_std]; this enables applications, such as embedded devices, which want to remove dependency code or provide their own core data structures. Internally, the standard library is divided into three parts, core, alloc, and std, where std and alloc are excluded by #![no_std].^[130]

Screenshot of crates.io in June 2022

Cargo

Cargo is Rust's build system and package manager. It downloads, compiles, distributes, and uploads packages—called crates—that are maintained in an official registry. It also acts as a front-end for Clippy and other Rust components.^[16]

By default, Cargo sources its dependencies from the user-contributed registry crates.io, but Git repositories and crates in the local filesystem, and other external sources can also be specified as dependencies.^[131]

Rustfmt

Rustfmt is a code formatter for Rust. It formats whitespace and indentation to produce code in accordance with a common style, unless otherwise specified. It can be invoked as a standalone program, or from a Rust project through Cargo.^[132]

Example output of Clippy on a hello world Rust program

Clippy

Clippy is Rust's built-in linting tool to improve the correctness, performance, and readability of Rust code. It was created in 2014^[133] and named after Microsoft Office's assistant, an anthropomorphized paperclip of the same name.^[134] As of 2024, it has more than 700 rules,^[135][136] which can be browsed online and filtered by category.^[137][138]

Versioning system

Following Rust 1.0, new features are developed in nightly versions which are released daily. During each six-week release cycle, changes to nightly versions are released to beta, while changes from the previous beta version are released to a new stable version.^[139]

Every two or three years, a new "edition" is produced. Editions are released to allow making limited breaking changes, such as promoting await to a keyword to support async/await features. Crates targeting different editions can interoperate with each other, so a crate can upgrade to a new edition even if its callers or its dependencies still target older editions. Migration to a new edition can be assisted with automated tooling.^[140]

IDE support

The most popular language server for Rust is Rust Analyzer, which officially replaced the original language server, RLS, in July 2022.^[141] Rust Analyzer provides IDEs and text editors with information about a Rust project; basic features including autocompletion, and the display of compilation errors while editing.^[142]

Rust Foundation

Rust Foundation

Formation	February 8, 2021 (2021-02-08)
Founders	Amazon Web Services Google Huawei Microsoft Mozilla Foundation
Type	Nonprofit organization
Location	United States
Chairperson	Shane Miller
Executive Director	Rebecca Rumbul
Website	foundation.rust-lang.org

The Rust Foundation is a non-profit membership organization incorporated in United States, with the primary purposes of backing the technical project as a legal entity and helping to manage the trademark and infrastructure assets.^[193][42]

It was established on February 8, 2021, with five founding corporate members (Amazon Web Services, Huawei, Google, Microsoft, and Mozilla).^[194] The foundation's board is chaired by Shane Miller.^[195] Starting in late 2021, its Executive Director and CEO is Rebecca Rumbul.^[196] Prior to this, Ashley Williams was interim executive director.^[197]

Governance teams

The Rust project is composed of teams that are responsible for different subareas of the development. The compiler team develops, manages, and optimizes compiler internals; and the language team designs new language features and helps implement them. The Rust project website lists 9 top-level teams as of January 2024.^[198] Representatives among teams form the Leadership council, which oversees the Rust project as a whole.^[199]

Book sources

Others