1 % The Rust Language Tutorial
5 Rust is a programming language with a focus on type safety, memory
6 safety, concurrency and performance. It is intended for writing
7 large-scale, high-performance software that is free from several
8 classes of common errors. Rust has a sophisticated memory model that
9 encourages efficient data structures and safe concurrency patterns,
10 forbidding invalid memory accesses that would otherwise cause
11 segmentation faults. It is statically typed and compiled ahead of
14 As a multi-paradigm language, Rust supports writing code in
15 procedural, functional and object-oriented styles. Some of its
16 pleasant high-level features include:
18 * **Type inference.** Type annotations on local variable declarations
20 * **Safe task-based concurrency.** Rust's lightweight tasks do not share
21 memory, instead communicating through messages.
22 * **Higher-order functions.** Efficient and flexible closures provide
23 iteration and other control structures
24 * **Pattern matching and algebraic data types.** Pattern matching on
25 Rust's enumeration types (a more powerful version of C's enums,
26 similar to algebraic data types in functional languages) is a
27 compact and expressive way to encode program logic.
28 * **Polymorphism.** Rust has type-parametric functions and
29 types, type classes and OO-style interfaces.
33 This is an introductory tutorial for the Rust programming language. It
34 covers the fundamentals of the language, including the syntax, the
35 type system and memory model, generics, and modules. [Additional
36 tutorials](#what-next?) cover specific language features in greater
39 This tutorial assumes that the reader is already familiar with one or
40 more languages in the C family. Understanding of pointers and general
41 memory management techniques will help.
45 Throughout the tutorial, language keywords and identifiers defined in
46 example code are displayed in `code font`.
48 Code snippets are indented, and also shown in a monospaced font. Not
49 all snippets constitute whole programs. For brevity, we'll often show
50 fragments of programs that don't compile on their own. To try them
51 out, you might have to wrap them in `fn main() { ... }`, and make sure
52 they don't contain references to names that aren't actually defined.
54 > *Warning:* Rust is a language under ongoing development. Notes
55 > about potential changes to the language, implementation
56 > deficiencies, and other caveats appear offset in blockquotes.
60 > *Warning:* The tarball and installer links are for the most recent
61 > release, not master. To use master, you **must** build from [git].
63 The Rust compiler currently must be built from a [tarball] or [git], unless
64 you are on Windows, in which case using the [installer][win-exe] is
65 recommended. There is a list of community-maintained nightly builds and
66 packages [on the wiki][wiki-packages].
68 Since the Rust compiler is written in Rust, it must be built by
69 a precompiled "snapshot" version of itself (made in an earlier state
70 of development). The source build automatically fetches these snapshots
71 from the Internet on our supported platforms.
73 Snapshot binaries are currently built and tested on several platforms:
75 * Windows (7, 8, Server 2008 R2), x86 only
76 * Linux (2.6.18 or later, various distributions), x86 and x86-64
77 * OSX 10.7 (Lion) or greater, x86 and x86-64
79 You may find that other platforms work, but these are our "tier 1"
80 supported build environments that are most likely to work.
82 > *Note:* Windows users should read the detailed
83 > [Getting started][wiki-start] notes on the wiki. Even when using
84 > the binary installer, the Windows build requires a MinGW installation,
85 > the precise details of which are not discussed here.
87 [wiki-start]: https://github.com/mozilla/rust/wiki/Note-getting-started-developing-Rust
88 [git]: https://github.com/mozilla/rust.git
90 To build from source you will also need the following prerequisite
93 * g++ 4.7 or clang++ 3.x
94 * python 2.6 or later (but not 3.x)
96 * gnu make 3.81 or later
99 If you've fulfilled those prerequisites, something along these lines
103 $ curl -O http://static.rust-lang.org/dist/rust-nightly.tar.gz
104 $ tar -xzf rust-nightly.tar.gz
107 $ make && make install
110 You may need to use `sudo make install` if you do not normally have
111 permission to modify the destination directory. The install locations
112 can be adjusted by passing a `--prefix` argument to
113 `configure`. Various other options are also supported: pass `--help`
114 for more information on them.
116 When complete, `make install` will place several programs into
117 `/usr/local/bin`: `rustc`, the Rust compiler, and `rustdoc`, the
118 API-documentation tool.
120 [tarball]: http://static.rust-lang.org/dist/rust-nightly.tar.gz
121 [win-exe]: http://static.rust-lang.org/dist/rust-nightly-install.exe
123 ## Compiling your first program
125 Rust program files are, by convention, given the extension `.rs`. Say
126 we have a file `hello.rs` containing this program:
133 > *Note:* An identifier followed by an exclamation point, like
134 > `println!`, is a macro invocation. Macros are explained
135 > [later](#syntax-extensions); for now just remember to include the
138 If the Rust compiler was installed successfully, running `rustc
139 hello.rs` will produce an executable called `hello` (or `hello.exe` on
140 Windows) which, upon running, will likely do exactly what you expect.
142 The Rust compiler tries to provide useful information when it encounters an
143 error. If you introduce an error into the program (for example, by changing
144 `println!` to some nonexistent macro), and then compile it, you'll see
145 an error message like this:
148 hello.rs:2:5: 2:24 error: macro undefined: 'print_with_unicorns'
149 hello.rs:2 print_with_unicorns!("hello?");
153 In its simplest form, a Rust program is a `.rs` file with some types
154 and functions defined in it. If it has a `main` function, it can be
155 compiled to an executable. Rust does not allow code that's not a
156 declaration to appear at the top level of the file: all statements must
157 live inside a function. Rust programs can also be compiled as
158 libraries, and included in other programs, even ones not written in Rust.
162 There are vim highlighting and indentation scripts in the Rust source
163 distribution under `src/etc/vim/`. There is an emacs mode under
164 `src/etc/emacs/` called `rust-mode`, but do read the instructions
165 included in that directory. In particular, if you are running emacs
166 24, then using emacs's internal package manager to install `rust-mode`
167 is the easiest way to keep it up to date. There is also a package for
168 Sublime Text 2, available both [standalone][sublime] and through
169 [Sublime Package Control][sublime-pkg], and support for Kate
170 under `src/etc/kate`.
172 A community-maintained list of available Rust tooling is [on the
173 wiki][wiki-packages].
175 There is ctags support via `src/etc/ctags.rust`, but many other
176 tools and editors are not yet supported. If you end up writing a Rust
177 mode for your favorite editor, let us know so that we can link to it.
179 [sublime]: http://github.com/dbp/sublime-rust
180 [sublime-pkg]: http://wbond.net/sublime_packages/package_control
184 Assuming you've programmed in any C-family language (C++, Java,
185 JavaScript, C#, or PHP), Rust will feel familiar. Code is arranged
186 in blocks delineated by curly braces; there are control structures
187 for branching and looping, like the familiar `if` and `while`; function
188 calls are written `myfunc(arg1, arg2)`; operators are written the same
189 and mostly have the same precedence as in C; comments are again like C;
190 module names are separated with double-colon (`::`) as with C++.
192 The main surface difference to be aware of is that the condition at
193 the head of control structures like `if` and `while` does not require
194 parentheses, while their bodies *must* be wrapped in
195 braces. Single-statement, unbraced bodies are not allowed.
198 # mod universe { pub fn recalibrate() -> bool { true } }
202 // A tricky calculation
203 if universe::recalibrate() {
210 The `let` keyword introduces a local variable. Variables are immutable by
211 default. To introduce a local variable that you can re-assign later, use `let
219 println!("count is {}", count);
224 Although Rust can almost always infer the types of local variables, you can
225 specify a variable's type by following it in the `let` with a colon, then the
226 type name. Static items, on the other hand, always require a type annotation.
230 static MONSTER_FACTOR: f64 = 57.8;
231 let monster_size = MONSTER_FACTOR * 10.0;
232 let monster_size: int = 50;
235 Local variables may shadow earlier declarations, as in the previous example:
236 `monster_size` was first declared as a `f64`, and then a second
237 `monster_size` was declared as an `int`. If you were to actually compile this
238 example, though, the compiler would determine that the first `monster_size` is
239 unused and issue a warning (because this situation is likely to indicate a
240 programmer error). For occasions where unused variables are intentional, their
241 names may be prefixed with an underscore to silence the warning, like `let
242 _monster_size = 50;`.
244 Rust identifiers start with an alphabetic
245 character or an underscore, and after that may contain any sequence of
246 alphabetic characters, numbers, or underscores. The preferred style is to
247 write function, variable, and module names with lowercase letters, using
248 underscores where they help readability, while writing types in camel case.
251 let my_variable = 100;
252 type MyType = int; // primitive types are _not_ camel case
255 ## Expressions and semicolons
257 Though it isn't apparent in all code, there is a fundamental
258 difference between Rust's syntax and predecessors like C.
259 Many constructs that are statements in C are expressions
260 in Rust, allowing code to be more concise. For example, you might
261 write a piece of code like this:
264 # let item = "salad";
268 } else if item == "muffin" {
275 But, in Rust, you don't have to repeat the name `price`:
278 # let item = "salad";
282 } else if item == "muffin" {
289 Both pieces of code are exactly equivalent: they assign a value to
290 `price` depending on the condition that holds. Note that there
291 are no semicolons in the blocks of the second snippet. This is
292 important: the lack of a semicolon after the last statement in a
293 braced block gives the whole block the value of that last expression.
295 Put another way, the semicolon in Rust *ignores the value of an expression*.
296 Thus, if the branches of the `if` had looked like `{ 4; }`, the above example
297 would simply assign `()` (unit or void) to `price`. But without the semicolon, each
298 branch has a different value, and `price` gets the value of the branch that
301 In short, everything that's not a declaration (declarations are `let` for
302 variables; `fn` for functions; and any top-level named items such as
303 [traits](#traits), [enum types](#enums), and static items) is an
304 expression, including function bodies.
307 fn is_four(x: int) -> bool {
308 // No need for a return statement. The result of the expression
309 // is used as the return value.
314 ## Primitive types and literals
316 There are general signed and unsigned integer types, `int` and `uint`,
317 as well as 8-, 16-, 32-, and 64-bit variants, `i8`, `u16`, etc.
318 Integers can be written in decimal (`144`), hexadecimal (`0x90`), octal (`0o70`), or
319 binary (`0b10010000`) base. Each integral type has a corresponding literal
320 suffix that can be used to indicate the type of a literal: `i` for `int`,
321 `u` for `uint`, `i8` for the `i8` type.
323 In the absence of an integer literal suffix, Rust will infer the
324 integer type based on type annotations and function signatures in the
325 surrounding program. In the absence of any type information at all,
326 Rust will assume that an unsuffixed integer literal has type
330 let a = 1; // `a` is an `int`
331 let b = 10i; // `b` is an `int`, due to the `i` suffix
332 let c = 100u; // `c` is a `uint`
333 let d = 1000i32; // `d` is an `i32`
336 There are two floating-point types: `f32`, and `f64`.
337 Floating-point numbers are written `0.0`, `1e6`, or `2.1e-4`.
338 Like integers, floating-point literals are inferred to the correct type.
339 Suffixes `f32`, and `f64` can be used to create literals of a specific type.
341 The keywords `true` and `false` produce literals of type `bool`.
343 Characters, the `char` type, are four-byte Unicode codepoints,
344 whose literals are written between single quotes, as in `'x'`.
345 Just like C, Rust understands a number of character escapes, using the backslash
346 character, such as `\n`, `\r`, and `\t`. String literals,
347 written between double quotes, allow the same escape sequences, and do no
348 other processing, unlike languages such as PHP or shell.
350 On the other hand, raw string literals do not process any escape sequences.
351 They are written as `r##"blah"##`, with a matching number of zero or more `#`
352 before the opening and after the closing quote, and can contain any sequence of
353 characters except their closing delimiter. More on strings
354 [later](#vectors-and-strings).
356 The unit type, written `()`, has a single value, also written `()`.
360 Rust's set of operators contains very few surprises. Arithmetic is done with
361 `*`, `/`, `%`, `+`, and `-` (multiply, quotient, remainder, add, and subtract). `-` is
362 also a unary prefix operator that negates numbers. As in C, the bitwise operators
363 `>>`, `<<`, `&`, `|`, and `^` are also supported.
365 Note that, if applied to an integer value, `!` flips all the bits (bitwise
368 The comparison operators are the traditional `==`, `!=`, `<`, `>`,
369 `<=`, and `>=`. Short-circuiting (lazy) boolean operators are written
370 `&&` (and) and `||` (or).
372 For compile-time type casting, Rust uses the binary `as` operator. It takes
373 an expression on the left side and a type on the right side and will, if a
374 meaningful conversion exists, convert the result of the expression to the
375 given type. Generally, `as` is only used with the primitive numeric types or
376 pointers, and is not overloadable. [`transmute`][transmute] can be used for
377 unsafe C-like casting of same-sized types.
381 let y: uint = x as uint;
385 [transmute]: http://static.rust-lang.org/doc/master/std/cast/fn.transmute.html
389 *Syntax extensions* are special forms that are not built into the language,
390 but are instead provided by the libraries. To make it clear to the reader when
391 a name refers to a syntax extension, the names of all syntax extensions end
392 with `!`. The standard library defines a few syntax extensions, the most
393 useful of which is [`format!`][fmt], a `sprintf`-like text formatter that you
394 will often see in examples, and its related family of macros: `print!`,
395 `println!`, and `write!`.
397 `format!` draws syntax from Python, but contains many of the same principles
398 that [printf][pf] has. Unlike printf, `format!` will give you a compile-time
399 error when the types of the directives don't match the types of the arguments.
402 # let mystery_object = ();
404 // `{}` will print the "default format" of a type
405 println!("{} is {}", "the answer", 43);
407 // `{:?}` will conveniently print any type
408 println!("what is this thing: {:?}", mystery_object);
411 [pf]: http://en.cppreference.com/w/cpp/io/c/fprintf
412 [fmt]: http://static.rust-lang.org/doc/master/std/fmt/index.html
414 You can define your own syntax extensions with the macro system. For details,
415 see the [macro tutorial][macros]. Note that macro definition is currently
416 considered an unstable feature.
422 We've seen `if` expressions a few times already. To recap, braces are
423 compulsory, an `if` can have an optional `else` clause, and multiple
424 `if`/`else` constructs can be chained together:
428 println!("that's odd");
432 println!("neither true nor false");
436 The condition given to an `if` construct *must* be of type `bool` (no
437 implicit conversion happens). If the arms are blocks that have a
438 value, this value must be of the same type for every arm in which
439 control reaches the end of the block:
442 fn signum(x: int) -> int {
451 Rust's `match` construct is a generalized, cleaned-up version of C's
452 `switch` construct. You provide it with a value and a number of
453 *arms*, each labelled with a pattern, and the code compares the value
454 against each pattern in order until one matches. The matching pattern
455 executes its corresponding arm.
460 0 => println!("zero"),
461 1 | 2 => println!("one or two"),
462 3..10 => println!("three to ten"),
463 _ => println!("something else")
467 Unlike in C, there is no "falling through" between arms: only one arm
468 executes, and it doesn't have to explicitly `break` out of the
469 construct when it is finished.
471 A `match` arm consists of a *pattern*, then a fat arrow `=>`, followed
472 by an *action* (expression). Each case is separated by commas. It is
473 often convenient to use a block expression for each case, in which case
474 the commas are optional as shown below. Literals are valid patterns and
475 match only their own value. A single arm may match multiple different
476 patterns by combining them with the pipe operator (`|`), so long as every
477 pattern binds the same set of variables. Ranges of numeric literal
478 patterns can be expressed with two dots, as in `M..N`. The underscore
479 (`_`) is a wildcard pattern that matches any single value. (`..`) is a
480 different wildcard that can match one or more fields in an `enum` variant.
485 0 => { println!("zero") }
486 _ => { println!("something else") }
490 `match` constructs must be *exhaustive*: they must have an arm
491 covering every possible case. For example, the typechecker would
492 reject the previous example if the arm with the wildcard pattern was
495 A powerful application of pattern matching is *destructuring*:
496 matching in order to bind names to the contents of data types.
498 > *Note:* The following code makes use of tuples (`(f64, f64)`) which
499 > are explained in section 5.3. For now you can think of tuples as a list of
504 fn angle(vector: (f64, f64)) -> f64 {
505 let pi = f64::consts::PI;
507 (0.0, y) if y < 0.0 => 1.5 * pi,
508 (0.0, _) => 0.5 * pi,
509 (x, y) => (y / x).atan()
514 A variable name in a pattern matches any value, *and* binds that name
515 to the value of the matched value inside of the arm's action. Thus, `(0.0,
516 y)` matches any tuple whose first element is zero, and binds `y` to
517 the second element. `(x, y)` matches any two-element tuple, and binds both
518 elements to variables. `(0.0,_)` matches any tuple whose first element is zero
519 and does not bind anything to the second element.
521 A subpattern can also be bound to a variable, using `variable @ pattern`. For
527 a @ 0..20 => println!("{} years old", a),
528 _ => println!("older than 21")
532 Any `match` arm can have a guard clause (written `if EXPR`), called a
533 *pattern guard*, which is an expression of type `bool` that
534 determines, after the pattern is found to match, whether the arm is
535 taken or not. The variables bound by the pattern are in scope in this
536 guard expression. The first arm in the `angle` example shows an
537 example of a pattern guard.
539 You've already seen simple `let` bindings, but `let` is a little
540 fancier than you've been led to believe. It, too, supports destructuring
541 patterns. For example, you can write this to extract the fields from a
542 tuple, introducing two variables at once: `a` and `b`.
545 # fn get_tuple_of_two_ints() -> (int, int) { (1, 1) }
546 let (a, b) = get_tuple_of_two_ints();
549 Let bindings only work with _irrefutable_ patterns: that is, patterns
550 that can never fail to match. This excludes `let` from matching
551 literals and most `enum` variants.
555 `while` denotes a loop that iterates as long as its given condition
556 (which must have type `bool`) evaluates to `true`. Inside a loop, the
557 keyword `break` aborts the loop, and `continue` aborts the current
558 iteration and continues with the next.
561 let mut cake_amount = 8;
562 while cake_amount > 0 {
567 `loop` denotes an infinite loop, and is the preferred way of writing `while true`:
573 if x % 5 == 0 { break; }
578 This code prints out a weird sequence of numbers and stops as soon as
579 it finds one that can be divided by five.
581 There is also a for-loop that can be used to iterate over a range of numbers:
584 for n in range(0, 5) {
589 The snippet above prints integer numbers under 5 starting at 0.
591 More generally, a for loop works with anything implementing the `Iterator` trait.
592 Data structures can provide one or more methods that return iterators over
593 their contents. For example, strings support iteration over their contents in
603 The snippet above prints the characters in "Hello" vertically, adding a new
604 line after each character.
611 Rust struct types must be declared before they are used using the `struct`
612 syntax: `struct Name { field1: T1, field2: T2 [, ...] }`, where `T1`, `T2`,
613 ... denote types. To construct a struct, use the same syntax, but leave off
614 the `struct`: for example: `Point { x: 1.0, y: 2.0 }`.
616 Structs are quite similar to C structs and are even laid out the same way in
617 memory (so you can read from a Rust struct in C, and vice-versa). Use the dot
618 operator to access struct fields, as in `mypoint.x`.
627 Structs have "inherited mutability", which means that any field of a struct
628 may be mutable, if the struct is in a mutable slot.
630 With a value (say, `mypoint`) of such a type in a mutable location, you can do
631 `mypoint.y += 1.0`. But in an immutable location, such an assignment to a
632 struct without inherited mutability would result in a type error.
635 # struct Point { x: f64, y: f64 }
636 let mut mypoint = Point { x: 1.0, y: 1.0 };
637 let origin = Point { x: 0.0, y: 0.0 };
639 mypoint.y += 1.0; // `mypoint` is mutable, and its fields as well
640 origin.y += 1.0; // ERROR: assigning to immutable field
643 `match` patterns destructure structs. The basic syntax is
644 `Name { fieldname: pattern, ... }`:
647 # struct Point { x: f64, y: f64 }
648 # let mypoint = Point { x: 0.0, y: 0.0 };
650 Point { x: 0.0, y: yy } => println!("{}", yy),
651 Point { x: xx, y: yy } => println!("{} {}", xx, yy)
655 In general, the field names of a struct do not have to appear in the same
656 order they appear in the type. When you are not interested in all
657 the fields of a struct, a struct pattern may end with `, ..` (as in
658 `Name { field1, .. }`) to indicate that you're ignoring all other fields.
659 Additionally, struct fields have a shorthand matching form that simply
660 reuses the field name as the binding name.
663 # struct Point { x: f64, y: f64 }
664 # let mypoint = Point { x: 0.0, y: 0.0 };
666 Point { x, .. } => println!("{}", x)
672 Enums are datatypes with several alternate representations. A simple `enum`
673 defines one or more constants, all of which have the same type:
684 Each variant of this enum has a unique and constant integral discriminator
685 value. If no explicit discriminator is specified for a variant, the value
686 defaults to the value of the previous variant plus one. If the first variant
687 does not have a discriminator, it defaults to 0. For example, the value of
688 `North` is 0, `East` is 1, `South` is 2, and `West` is 3.
690 When an enum has simple integer discriminators, you can apply the `as` cast
691 operator to convert a variant to its discriminator value as an `int`:
694 # enum Direction { North }
695 println!( "{:?} => {}", North, North as int );
698 It is possible to set the discriminator values to chosen constant values:
708 Variants do not have to be simple values; they may be more complex:
711 # struct Point { x: f64, y: f64 }
714 Rectangle(Point, Point)
718 A value of this type is either a `Circle`, in which case it contains a
719 `Point` struct and a f64, or a `Rectangle`, in which case it contains
720 two `Point` structs. The run-time representation of such a value
721 includes an identifier of the actual form that it holds, much like the
722 "tagged union" pattern in C, but with better static guarantees.
724 This declaration defines a type `Shape` that can refer to such shapes, and two
725 functions, `Circle` and `Rectangle`, which can be used to construct values of
726 the type. To create a new Circle, write `Circle(Point { x: 0.0, y: 0.0 },
729 All of these variant constructors may be used as patterns. The only way to
730 access the contents of an enum instance is the destructuring of a match. For
735 # struct Point {x: f64, y: f64}
736 # enum Shape { Circle(Point, f64), Rectangle(Point, Point) }
737 fn area(sh: Shape) -> f64 {
739 Circle(_, size) => f64::consts::PI * size * size,
740 Rectangle(Point { x, y }, Point { x: x2, y: y2 }) => (x2 - x) * (y2 - y)
745 Use a lone `_` to ignore an individual field. Ignore all fields of a variant
746 like: `Circle(..)`. Nullary enum patterns are written without parentheses:
749 # struct Point { x: f64, y: f64 }
750 # enum Direction { North, East, South, West }
751 fn point_from_direction(dir: Direction) -> Point {
753 North => Point { x: 0.0, y: 1.0 },
754 East => Point { x: 1.0, y: 0.0 },
755 South => Point { x: 0.0, y: -1.0 },
756 West => Point { x: -1.0, y: 0.0 }
761 Enum variants may also be structs. For example:
765 # struct Point { x: f64, y: f64 }
766 # fn square(x: f64) -> f64 { x * x }
768 Circle { center: Point, radius: f64 },
769 Rectangle { top_left: Point, bottom_right: Point }
771 fn area(sh: Shape) -> f64 {
773 Circle { radius: radius, .. } => f64::consts::PI * square(radius),
774 Rectangle { top_left: top_left, bottom_right: bottom_right } => {
775 (bottom_right.x - top_left.x) * (top_left.y - bottom_right.y)
781 > *Note:* This feature of the compiler is currently gated behind the
782 > `#[feature(struct_variant)]` directive. More about these directives can be
783 > found in the manual.
787 Tuples in Rust behave exactly like structs, except that their fields do not
788 have names. Thus, you cannot access their fields with dot notation. Tuples
789 can have any arity (number of elements) except for 0 (though you may consider
790 unit, `()`, as the empty tuple if you like).
793 let mytup: (int, int, f64) = (10, 20, 30.0);
795 (a, b, c) => println!("{}", a + b + (c as int))
801 Rust also has _tuple structs_, which behave like both structs and tuples,
802 except that, unlike tuples, tuple structs have names (so `Foo(1, 2)` has a
803 different type from `Bar(1, 2)`), and tuple structs' _fields_ do not have
809 struct MyTup(int, int, f64);
810 let mytup: MyTup = MyTup(10, 20, 30.0);
812 MyTup(a, b, c) => println!("{}", a + b + (c as int))
816 <a name="newtype"></a>
818 There is a special case for tuple structs with a single field, which are
819 sometimes called "newtypes" (after Haskell's "newtype" feature). These are
820 used to define new types in such a way that the new name is not just a
821 synonym for an existing type but is rather its own distinct type.
827 Types like this can be useful to differentiate between data that have
828 the same underlying type but must be used in different ways.
832 struct Centimeters(int);
835 The above definitions allow for a simple way for programs to avoid
836 confusing numbers that correspond to different units. Their integer
837 values can be extracted with pattern matching:
840 # struct Inches(int);
841 let length_with_unit = Inches(10);
842 let Inches(integer_length) = length_with_unit;
843 println!("length is {} inches", integer_length);
848 We've already seen several function definitions. Like all other static
849 declarations, such as `type`, functions can be declared both at the
850 top level and inside other functions (or in modules, which we'll come
851 back to [later](#crates-and-the-module-system)). The `fn` keyword introduces a
852 function. A function has an argument list, which is a parenthesized
853 list of `name: type` pairs separated by commas. An arrow `->`
854 separates the argument list and the function's return type.
857 fn line(a: int, b: int, x: int) -> int {
862 The `return` keyword immediately returns from the body of a function. It
863 is optionally followed by an expression to return. A function can
864 also return a value by having its top-level block produce an
868 fn line(a: int, b: int, x: int) -> int {
873 It's better Rust style to write a return value this way instead of
874 writing an explicit `return`. The utility of `return` comes in when
875 returning early from a function. Functions that do not return a value
876 are said to return unit, `()`, and both the return type and the return
877 value may be omitted from the definition. The following two functions
881 fn do_nothing_the_hard_way() -> () { return (); }
883 fn do_nothing_the_easy_way() { }
886 Ending the function with a semicolon like so is equivalent to returning `()`.
889 fn line(a: int, b: int, x: int) -> int { a * x + b }
890 fn oops(a: int, b: int, x: int) -> () { a * x + b; }
892 assert!(8 == line(5, 3, 1));
893 assert!(() == oops(5, 3, 1));
896 As with `match` expressions and `let` bindings, function arguments support
897 pattern destructuring. Like `let`, argument patterns must be irrefutable,
898 as in this example that unpacks the first value from a tuple and returns it.
901 fn first((value, _): (int, f64)) -> int { value }
906 A *destructor* is a function responsible for cleaning up the resources used by
907 an object when it is no longer accessible. Destructors can be defined to handle
908 the release of resources like files, sockets and heap memory.
910 Objects are never accessible after their destructor has been called, so no
911 dynamic failures are possible from accessing freed resources. When a task
912 fails, destructors of all objects in the task are called.
914 The `~` sigil represents a unique handle for a memory allocation on the heap:
918 // an integer allocated on the heap
921 // the destructor frees the heap memory as soon as `y` goes out of scope
924 Rust includes syntax for heap memory allocation in the language since it's
925 commonly used, but the same semantics can be implemented by a type with a
930 Rust formalizes the concept of object ownership to delegate management of an
931 object's lifetime to either a variable or a task-local garbage collector. An
932 object's owner is responsible for managing the lifetime of the object by
933 calling the destructor, and the owner determines whether the object is mutable.
935 Ownership is recursive, so mutability is inherited recursively and a destructor
936 destroys the contained tree of owned objects. Variables are top-level owners
937 and destroy the contained object when they go out of scope.
940 // the struct owns the objects contained in the `x` and `y` fields
941 struct Foo { x: int, y: ~int }
944 // `a` is the owner of the struct, and thus the owner of the struct's fields
945 let a = Foo { x: 5, y: ~10 };
947 // when `a` goes out of scope, the destructor for the `~int` in the struct's
950 // `b` is mutable, and the mutability is inherited by the objects it owns
951 let mut b = Foo { x: 5, y: ~10 };
955 If an object doesn't contain any non-Send types, it consists of a single
956 ownership tree and is itself given the `Send` trait which allows it to be sent
957 between tasks. Custom destructors can only be implemented directly on types
958 that are `Send`, but non-`Send` types can still *contain* types with custom
959 destructors. Example of types which are not `Send` are [`Gc<T>`][gc] and
960 [`Rc<T>`][rc], the shared-ownership types.
962 [gc]: http://static.rust-lang.org/doc/master/std/gc/struct.Gc.html
963 [rc]: http://static.rust-lang.org/doc/master/std/rc/struct.Rc.html
965 # Implementing a linked list
967 An `enum` is a natural fit for describing a linked list, because it can express
968 a `List` type as being *either* the end of the list (`Nil`) or another node
969 (`Cons`). The full definition of the `Cons` variant will require some thought.
973 Cons(...), // an incomplete definition of the next element in a List
974 Nil // the end of a List
978 The obvious approach is to define `Cons` as containing an element in the list
979 along with the next `List` node. However, this will generate a compiler error.
982 // error: illegal recursive enum type; wrap the inner value in a box to make it
985 Cons(u32, List), // an element (`u32`) and the next node in the list
990 This error message is related to Rust's precise control over memory layout, and
991 solving it will require introducing the concept of *boxing*.
995 A value in Rust is stored directly inside the owner. If a `struct` contains
996 four `u32` fields, it will be four times as large as a single `u32`.
999 use std::mem::size_of; // bring `size_of` into the current scope, for convenience
1008 assert_eq!(size_of::<Foo>(), size_of::<u32>() * 4);
1017 assert_eq!(size_of::<Bar>(), size_of::<u32>() * 16);
1020 Our previous attempt at defining the `List` type included an `u32` and a `List`
1021 directly inside `Cons`, making it at least as big as the sum of both types. The
1022 type was invalid because the size was infinite!
1024 An *owned box* (`~`) uses a dynamic memory allocation to provide the invariant
1025 of always being the size of a pointer, regardless of the contained type. This
1026 can be leveraged to create a valid `List` definition:
1035 Defining a recursive data structure like this is the canonical example of an
1036 owned box. Much like an unboxed value, an owned box has a single owner and is
1037 therefore limited to expressing a tree-like data structure.
1039 Consider an instance of our `List` type:
1046 let list = Cons(1, ~Cons(2, ~Cons(3, ~Nil)));
1049 It represents an owned tree of values, inheriting mutability down the tree and
1050 being destroyed along with the owner. Since the `list` variable above is
1051 immutable, the whole list is immutable. The memory allocation itself is the
1052 box, while the owner holds onto a pointer to it:
1055 List box List box List box List box
1056 +--------------+ +--------------+ +--------------+ +----------+
1057 list -> | Cons | 1 | ~ | -> | Cons | 2 | ~ | -> | Cons | 3 | ~ | -> | Nil |
1058 +--------------+ +--------------+ +--------------+ +----------+
1061 > *Note:* the above diagram shows the logical contents of the enum. The actual
1062 > memory layout of the enum may vary. For example, for the `List` enum shown
1063 > above, Rust guarantees that there will be no enum tag field in the actual
1064 > structure. See the language reference for more details.
1066 An owned box is a common example of a type with a destructor. The allocated
1067 memory is cleaned up when the box is destroyed.
1071 Rust uses a shallow copy for parameter passing, assignment and returning from
1072 functions. Passing around the `List` will copy only as deep as the pointer to
1073 the box rather than doing an implicit heap allocation.
1080 let xs = Cons(1, ~Cons(2, ~Cons(3, ~Nil)));
1081 let ys = xs; // copies `Cons(u32, pointer)` shallowly
1084 Rust will consider a shallow copy of a type with a destructor like `List` to
1085 *move ownership* of the value. After a value has been moved, the source
1086 location cannot be used unless it is reinitialized.
1096 // attempting to use `xs` will result in an error here
1100 // `xs` can be used again
1103 A destructor call will only occur for a variable that has not been moved from,
1104 as it is only called a single time.
1107 Avoiding a move can be done with the library-defined `clone` method:
1111 let y = x.clone(); // `y` is a newly allocated box
1112 let z = x; // no new memory allocated, `x` can no longer be used
1115 The `clone` method is provided by the `Clone` trait, and can be derived for
1116 our `List` type. Traits will be explained in detail [later](#traits).
1125 let x = Cons(5, ~Nil);
1128 // `x` can still be used!
1132 // and now, it can no longer be used since it has been moved
1135 The mutability of a value may be changed by moving it to a new owner:
1139 let mut s = r; // box becomes mutable
1141 let t = s; // box becomes immutable
1144 A simple way to define a function prepending to the `List` type is to take
1153 fn prepend(xs: List, value: u32) -> List {
1158 xs = prepend(xs, 1);
1159 xs = prepend(xs, 2);
1160 xs = prepend(xs, 3);
1163 However, this is not a very flexible definition of `prepend` as it requires
1164 ownership of a list to be passed in rather than just mutating it in-place.
1168 The obvious signature for a `List` equality comparison is the following:
1171 fn eq(xs: List, ys: List) -> bool { /* ... */ }
1174 However, this will cause both lists to be moved into the function. Ownership
1175 isn't required to compare the lists, so the function should take *references*
1179 fn eq(xs: &List, ys: &List) -> bool { /* ... */ }
1182 A reference is a *non-owning* view of a value. A reference can be obtained with the `&` (address-of)
1183 operator. It can be dereferenced by using the `*` operator. In a pattern, such as `match` expression
1184 branches, the `ref` keyword can be used to bind to a variable name by-reference rather than
1185 by-value. A recursive definition of equality using references is as follows:
1192 fn eq(xs: &List, ys: &List) -> bool {
1193 // Match on the next node in both lists.
1195 // If we have reached the end of both lists, they are equal.
1196 (&Nil, &Nil) => true,
1197 // If the current elements of both lists are equal, keep going.
1198 (&Cons(x, ~ref next_xs), &Cons(y, ~ref next_ys))
1199 if x == y => eq(next_xs, next_ys),
1200 // If the current elements are not equal, the lists are not equal.
1205 let xs = Cons(5, ~Cons(10, ~Nil));
1206 let ys = Cons(5, ~Cons(10, ~Nil));
1207 assert!(eq(&xs, &ys));
1210 > *Note:* Rust doesn't guarantee [tail-call](http://en.wikipedia.org/wiki/Tail_call) optimization,
1211 > but LLVM is able to handle a simple case like this with optimizations enabled.
1213 ## Lists of other types
1215 Our `List` type is currently always a list of 32-bit unsigned integers. By
1216 leveraging Rust's support for generics, it can be extended to work for any
1219 The `u32` in the previous definition can be substituted with a type parameter:
1221 > *Note:* The following code introduces generics, which are explained in a
1222 > [dedicated section](#generics).
1231 The old `List` of `u32` is now available as `List<u32>`. The `prepend`
1232 definition has to be updated too:
1236 # Cons(T, ~List<T>),
1239 fn prepend<T>(xs: List<T>, value: T) -> List<T> {
1244 Generic functions and types like this are equivalent to defining specialized
1245 versions for each set of type parameters.
1247 Using the generic `List<T>` works much like before, thanks to type inference:
1251 # Cons(T, ~List<T>),
1254 # fn prepend<T>(xs: List<T>, value: T) -> List<T> {
1257 let mut xs = Nil; // Unknown type! This is a `List<T>`, but `T` can be anything.
1258 xs = prepend(xs, 10); // Here the compiler infers `xs`'s type as `List<int>`.
1259 xs = prepend(xs, 15);
1260 xs = prepend(xs, 20);
1263 The code sample above demonstrates type inference making most type annotations optional. It is
1264 equivalent to the following type-annotated code:
1268 # Cons(T, ~List<T>),
1271 # fn prepend<T>(xs: List<T>, value: T) -> List<T> {
1274 let mut xs: List<int> = Nil::<int>;
1275 xs = prepend::<int>(xs, 10);
1276 xs = prepend::<int>(xs, 15);
1277 xs = prepend::<int>(xs, 20);
1280 In declarations, the language uses `Type<T, U, V>` to describe a list of type
1281 parameters, but expressions use `identifier::<T, U, V>`, to disambiguate the
1284 ## Defining list equality with generics
1286 Generic functions are type-checked from the definition, so any necessary properties of the type must
1287 be specified up-front. Our previous definition of list equality relied on the element type having
1288 the `==` operator available, and took advantage of the lack of a destructor on `u32` to copy it
1289 without a move of ownership.
1291 We can add a *trait bound* on the `Eq` trait to require that the type implement the `==` operator.
1292 Two more `ref` annotations need to be added to avoid attempting to move out the element types:
1296 # Cons(T, ~List<T>),
1299 fn eq<T: Eq>(xs: &List<T>, ys: &List<T>) -> bool {
1300 // Match on the next node in both lists.
1302 // If we have reached the end of both lists, they are equal.
1303 (&Nil, &Nil) => true,
1304 // If the current elements of both lists are equal, keep going.
1305 (&Cons(ref x, ~ref next_xs), &Cons(ref y, ~ref next_ys))
1306 if x == y => eq(next_xs, next_ys),
1307 // If the current elements are not equal, the lists are not equal.
1312 let xs = Cons('c', ~Cons('a', ~Cons('t', ~Nil)));
1313 let ys = Cons('c', ~Cons('a', ~Cons('t', ~Nil)));
1314 assert!(eq(&xs, &ys));
1317 This would be a good opportunity to implement the `Eq` trait for our list type, making the `==` and
1318 `!=` operators available. We'll need to provide an `impl` for the `Eq` trait and a definition of the
1319 `eq` method. In a method, the `self` parameter refers to an instance of the type we're implementing
1324 # Cons(T, ~List<T>),
1327 impl<T: Eq> Eq for List<T> {
1328 fn eq(&self, ys: &List<T>) -> bool {
1329 // Match on the next node in both lists.
1331 // If we have reached the end of both lists, they are equal.
1332 (&Nil, &Nil) => true,
1333 // If the current elements of both lists are equal, keep going.
1334 (&Cons(ref x, ~ref next_xs), &Cons(ref y, ~ref next_ys))
1335 if x == y => next_xs == next_ys,
1336 // If the current elements are not equal, the lists are not equal.
1342 let xs = Cons(5, ~Cons(10, ~Nil));
1343 let ys = Cons(5, ~Cons(10, ~Nil));
1344 // The methods below are part of the Eq trait,
1345 // which we implemented on our linked list.
1346 assert!(xs.eq(&ys));
1347 assert!(!xs.ne(&ys));
1349 // The Eq trait also allows us to use the shorthand infix operators.
1350 assert!(xs == ys); // `xs == ys` is short for `xs.eq(&ys)`
1351 assert!(!(xs != ys)); // `xs != ys` is short for `xs.ne(&ys)`
1356 The most common use case for owned boxes is creating recursive data structures
1357 like a binary search tree. Rust's trait-based generics system (covered later in
1358 the tutorial) is usually used for static dispatch, but also provides dynamic
1359 dispatch via boxing. Values of different types may have different sizes, but a
1360 box is able to *erase* the difference via the layer of indirection they
1363 In uncommon cases, the indirection can provide a performance gain or memory
1364 reduction by making values smaller. However, unboxed values should almost
1365 always be preferred when they are usable.
1367 Note that returning large unboxed values via boxes is unnecessary. A large
1368 value is returned via a hidden output parameter, and the decision on where to
1369 place the return value should be left to the caller:
1372 fn foo() -> (u64, u64, u64, u64, u64, u64) {
1376 let x = ~foo(); // allocates a `~` box, and writes the integers directly to it
1379 Beyond the properties granted by the size, an owned box behaves as a regular
1380 value by inheriting the mutability and lifetime of the owner:
1383 let x = 5; // immutable
1384 let mut y = 5; // mutable
1387 let x = ~5; // immutable
1388 let mut y = ~5; // mutable
1389 *y += 2; // the `*` operator is needed to access the contained value
1395 owned boxes, where the holder of an owned box is the owner of the pointed-to
1396 memory, references never imply ownership - they are "borrowed".
1397 You can borrow a reference to
1398 any object, and the compiler verifies that it cannot outlive the lifetime of
1401 As an example, consider a simple struct type, `Point`:
1410 We can use this simple definition to allocate points in many different
1411 ways. For example, in this code, each of these local variables
1412 contains a point, but allocated in a different location:
1415 # struct Point { x: f64, y: f64 }
1416 let on_the_stack : Point = Point { x: 3.0, y: 4.0 };
1417 let owned_box : ~Point = ~Point { x: 7.0, y: 9.0 };
1420 Suppose we want to write a procedure that computes the distance
1421 between any two points, no matter where they are stored. One option is
1422 to define a function that takes two arguments of type point—that is,
1423 it takes the points by value. But this will cause the points to be
1424 copied when we call the function. For points, this is probably not so
1425 bad, but often copies are expensive. So we’d like to define a function
1426 that takes the points by pointer. We can use references to do this:
1429 # struct Point { x: f64, y: f64 }
1430 fn compute_distance(p1: &Point, p2: &Point) -> f64 {
1431 let x_d = p1.x - p2.x;
1432 let y_d = p1.y - p2.y;
1433 (x_d * x_d + y_d * y_d).sqrt()
1437 Now we can call `compute_distance()` in various ways:
1440 # struct Point{ x: f64, y: f64 };
1441 # let on_the_stack : Point = Point { x: 3.0, y: 4.0 };
1442 # let owned_box : ~Point = ~Point { x: 7.0, y: 9.0 };
1443 # fn compute_distance(p1: &Point, p2: &Point) -> f64 { 0.0 }
1444 compute_distance(&on_the_stack, owned_box);
1447 Here the `&` operator is used to take the address of the variable
1448 `on_the_stack`; this is because `on_the_stack` has the type `Point`
1449 (that is, a struct value) and we have to take its address to get a
1450 reference. We also call this _borrowing_ the local variable
1451 `on_the_stack`, because we are creating an alias: that is, another
1452 route to the same data.
1454 In the case of `owned_box`, however, no
1455 explicit action is necessary. The compiler will automatically convert
1456 a box `~point` to a reference like
1457 `&point`. This is another form of borrowing; in this case, the
1458 contents of the owned box are being lent out.
1460 Whenever a value is borrowed, there are some limitations on what you
1461 can do with the original. For example, if the contents of a variable
1462 have been lent out, you cannot send that variable to another task, nor
1463 will you be permitted to take actions that might cause the borrowed
1464 value to be freed or to change its type. This rule should make
1465 intuitive sense: you must wait for a borrowed value to be returned
1466 (that is, for the reference to go out of scope) before you can
1467 make full use of it again.
1469 For a more in-depth explanation of references and lifetimes, read the
1470 [references and lifetimes guide][lifetimes].
1474 Lending an &-pointer to an object freezes it and prevents mutation—even if the object was declared as `mut`.
1475 `Freeze` objects have freezing enforced statically at compile-time. An example
1476 of a non-`Freeze` type is [`RefCell<T>`][refcell].
1481 let y = &x; // `x` is now frozen. It cannot be modified or re-assigned.
1483 // `x` is now unfrozen again
1487 [refcell]: http://static.rust-lang.org/doc/master/std/cell/struct.RefCell.html
1489 # Dereferencing pointers
1491 Rust uses the unary star operator (`*`) to access the contents of a
1492 box or pointer, similarly to C.
1498 let sum = *owned + *borrowed;
1501 Dereferenced mutable pointers may appear on the left hand side of
1502 assignments. Such an assignment modifies the value that the pointer
1506 let mut owned = ~10;
1509 let borrowed = &mut value;
1511 *owned = *borrowed + 100;
1512 *borrowed = *owned + 1000;
1515 Pointers have high operator precedence, but lower precedence than the
1516 dot operator used for field and method access. This precedence order
1517 can sometimes make code awkward and parenthesis-filled.
1520 # struct Point { x: f64, y: f64 }
1521 # enum Shape { Rectangle(Point, Point) }
1522 # impl Shape { fn area(&self) -> int { 0 } }
1523 let start = ~Point { x: 10.0, y: 20.0 };
1524 let end = ~Point { x: (*start).x + 100.0, y: (*start).y + 100.0 };
1525 let rect = &Rectangle(*start, *end);
1526 let area = (*rect).area();
1529 To combat this ugliness the dot operator applies _automatic pointer
1530 dereferencing_ to the receiver (the value on the left-hand side of the
1531 dot), so in most cases, explicitly dereferencing the receiver is not necessary.
1534 # struct Point { x: f64, y: f64 }
1535 # enum Shape { Rectangle(Point, Point) }
1536 # impl Shape { fn area(&self) -> int { 0 } }
1537 let start = ~Point { x: 10.0, y: 20.0 };
1538 let end = ~Point { x: start.x + 100.0, y: start.y + 100.0 };
1539 let rect = &Rectangle(*start, *end);
1540 let area = rect.area();
1543 You can write an expression that dereferences any number of pointers
1544 automatically. For example, if you feel inclined, you could write
1545 something silly like
1548 # struct Point { x: f64, y: f64 }
1549 let point = &~Point { x: 10.0, y: 20.0 };
1550 println!("{:f}", point.x);
1553 The indexing operator (`[]`) also auto-dereferences.
1555 # Vectors and strings
1557 A vector is a contiguous block of memory containing zero or more values of the
1558 same type. Rust also supports vector reference types, called slices, which are
1559 a view into a block of memory represented as a pointer and a length.
1561 Strings are represented as vectors of `u8`, with the guarantee of containing a
1562 valid UTF-8 sequence.
1564 Fixed-size vectors are an unboxed block of memory, with the element length as
1565 part of the type. A fixed-size vector owns the elements it contains, so the
1566 elements are mutable if the vector is mutable. Fixed-size strings do not exist.
1569 // A fixed-size vector
1570 let numbers = [1, 2, 3];
1571 let more_numbers = numbers;
1573 // The type of a fixed-size vector is written as `[Type, ..length]`
1574 let five_zeroes: [int, ..5] = [0, ..5];
1577 A unique vector is dynamically sized, and has a destructor to clean up
1578 allocated memory on the heap. A unique vector owns the elements it contains, so
1579 the elements are mutable if the vector is mutable.
1582 use std::strbuf::StrBuf;
1584 // A dynamically sized vector (unique vector)
1585 let mut numbers = vec![1, 2, 3];
1589 // The type of a unique vector is written as `~[int]`
1590 let more_numbers: ~[int] = numbers.move_iter().collect();
1592 // The original `numbers` value can no longer be used, due to move semantics.
1594 let mut string = StrBuf::from_str("fo");
1595 string.push_char('o');
1598 Slices are similar to fixed-size vectors, but the length is not part of the
1599 type. They simply point into a block of memory and do not have ownership over
1604 let xs = &[1, 2, 3];
1606 // Slices have their type written as `&[int]`
1607 let ys: &[int] = xs;
1609 // Other vector types coerce to slices
1610 let three = [1, 2, 3];
1611 let zs: &[int] = three;
1613 // An unadorned string literal is an immutable string slice
1614 let string = "foobar";
1616 // A string slice type is written as `&str`
1617 let view: &str = string.slice(0, 3);
1620 Mutable slices also exist, just as there are mutable references. However, there
1621 are no mutable string slices. Strings are a multi-byte encoding (UTF-8) of
1622 Unicode code points, so they cannot be freely mutated without the ability to
1626 let mut xs = [1, 2, 3];
1627 let view = xs.mut_slice(0, 2);
1630 // The type of a mutable slice is written as `&mut [T]`
1631 let ys: &mut [int] = &mut [1, 2, 3];
1634 Square brackets denote indexing into a vector:
1637 # enum Crayon { Almond, AntiqueBrass, Apricot,
1638 # Aquamarine, Asparagus, AtomicTangerine,
1639 # BananaMania, Beaver, Bittersweet };
1640 # fn draw_scene(c: Crayon) { }
1641 let crayons: [Crayon, ..3] = [BananaMania, Beaver, Bittersweet];
1643 Bittersweet => draw_scene(crayons[0]),
1648 A vector can be destructured using pattern matching:
1651 let numbers: &[int] = &[1, 2, 3];
1652 let score = match numbers {
1655 [a, b] => a * 6 + b * 4,
1656 [a, b, c, ..rest] => a * 5 + b * 3 + c * 2 + rest.len() as int
1660 Both vectors and strings support a number of useful [methods](#methods),
1661 defined in [`std::vec`] and [`std::str`].
1663 [`std::vec`]: std/vec/index.html
1664 [`std::str`]: std/str/index.html
1666 # Ownership escape hatches
1668 Ownership can cleanly describe tree-like data structures, and references provide non-owning pointers. However, more flexibility is often desired and Rust provides ways to escape from strict
1669 single parent ownership.
1671 The standard library provides the `std::rc::Rc` pointer type to express *shared ownership* over a
1672 reference counted box. As soon as all of the `Rc` pointers go out of scope, the box and the
1673 contained value are destroyed.
1678 // A fixed-size array allocated in a reference-counted box
1679 let x = Rc::new([1, 2, 3, 4, 5, 6, 7, 8, 9, 10]);
1680 let y = x.clone(); // a new owner
1681 let z = x; // this moves `x` into `z`, rather than creating a new owner
1683 assert!(*z == [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]);
1685 // the variable is mutable, but not the contents of the box
1686 let mut a = Rc::new([10, 9, 8, 7, 6, 5, 4, 3, 2, 1]);
1690 A garbage collected pointer is provided via `std::gc::Gc`, with a task-local garbage collector
1691 having ownership of the box. It allows the creation of cycles, and the individual `Gc` pointers do
1692 not have a destructor.
1697 // A fixed-size array allocated in a garbage-collected box
1698 let x = Gc::new([1, 2, 3, 4, 5, 6, 7, 8, 9, 10]);
1699 let y = x; // does not perform a move, unlike with `Rc`
1702 assert!(*z.borrow() == [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]);
1705 With shared ownership, mutability cannot be inherited so the boxes are always immutable. However,
1706 it's possible to use *dynamic* mutability via types like `std::cell::Cell` where freezing is handled
1707 via dynamic checks and can fail at runtime.
1709 The `Rc` and `Gc` types are not sendable, so they cannot be used to share memory between tasks. Safe
1710 immutable and mutable shared memory is provided by the `extra::arc` module.
1714 Named functions, like those we've seen so far, may not refer to local
1715 variables declared outside the function: they do not close over their
1716 environment (sometimes referred to as "capturing" variables in their
1717 environment). For example, you couldn't write the following:
1722 // `fun` cannot refer to `x`
1723 fn fun() -> () { println!("{}", x); }
1726 A _closure_ does support accessing the enclosing scope; below we will create
1727 2 _closures_ (nameless functions). Compare how `||` replaces `()` and how
1728 they try to access `x`:
1733 // `fun` is an invalid definition
1734 fn fun () -> () { println!("{}", x) } // cannot capture from enclosing scope
1735 let closure = || -> () { println!("{}", x) }; // can capture from enclosing scope
1737 // `fun_arg` is an invalid definition
1738 fn fun_arg (arg: int) -> () { println!("{}", arg + x) } // cannot capture
1739 let closure_arg = |arg: int| -> () { println!("{}", arg + x) }; // can capture
1741 // Requires a type because the implementation needs to know which `+` to use.
1742 // In the future, the implementation may not need the help.
1744 fun(); // Still won't work
1745 closure(); // Prints: 3
1747 fun_arg(7); // Still won't work
1748 closure_arg(7); // Prints: 10
1751 Closures begin with the argument list between vertical bars and are followed by
1752 a single expression. Remember that a block, `{ <expr1>; <expr2>; ... }`, is
1753 considered a single expression: it evaluates to the result of the last
1754 expression it contains if that expression is not followed by a semicolon,
1755 otherwise the block evaluates to `()`, the unit value.
1757 In general, return types and all argument types must be specified
1758 explicitly for function definitions. (As previously mentioned in the
1759 [Functions section](#functions), omitting the return type from a
1760 function declaration is synonymous with an explicit declaration of
1761 return type unit, `()`.)
1764 fn fun (x: int) { println!("{}", x) } // this is same as saying `-> ()`
1765 fn square(x: int) -> uint { (x * x) as uint } // other return types are explicit
1767 // Error: mismatched types: expected `()` but found `uint`
1768 fn badfun(x: int) { (x * x) as uint }
1771 On the other hand, the compiler can usually infer both the argument
1772 and return types for a closure expression; therefore they are often
1773 omitted, since both a human reader and the compiler can deduce the
1774 types from the immediate context. This is in contrast to function
1775 declarations, which require types to be specified and are not subject
1776 to type inference. Compare:
1779 // `fun` as a function declaration cannot infer the type of `x`, so it must be provided
1780 fn fun (x: int) { println!("{}", x) }
1781 let closure = |x | { println!("{}", x) }; // infers `x: int`, return type `()`
1783 // For closures, omitting a return type is *not* synonymous with `-> ()`
1784 let add_3 = |y | { 3i + y }; // infers `y: int`, return type `int`.
1786 fun(10); // Prints 10
1787 closure(20); // Prints 20
1788 closure(add_3(30)); // Prints 33
1790 fun("String"); // Error: mismatched types
1792 // Error: mismatched types
1793 // inference already assigned `closure` the type `|int| -> ()`
1797 In cases where the compiler needs assistance, the arguments and return
1798 types may be annotated on closures, using the same notation as shown
1799 earlier. In the example below, since different types provide an
1800 implementation for the operator `*`, the argument type for the `x`
1801 parameter must be explicitly provided.
1804 // Error: the type of `x` must be known to be used with `x * x`
1805 let square = |x | -> uint { (x * x) as uint };
1808 In the corrected version, the argument type is explicitly annotated,
1809 while the return type can still be inferred.
1812 let square_explicit = |x: int| -> uint { (x * x) as uint };
1813 let square_infer = |x: int| { (x * x) as uint };
1815 println!("{}", square_explicit(20)); // 400
1816 println!("{}", square_infer(-20)); // 400
1819 There are several forms of closure, each with its own role. The most
1820 common, called a _stack closure_, has type `||` and can directly
1821 access local variables in the enclosing scope.
1825 let f = |x: int| if x > max { max = x };
1826 for x in [1, 2, 3].iter() {
1831 Stack closures are very efficient because their environment is
1832 allocated on the call stack and refers by pointer to captured
1833 locals. To ensure that stack closures never outlive the local
1834 variables to which they refer, stack closures are not
1835 first-class. That is, they can only be used in argument position; they
1836 cannot be stored in data structures or returned from
1837 functions. Despite these limitations, stack closures are used
1838 pervasively in Rust code.
1842 Owned closures, written `proc`,
1843 hold on to things that can safely be sent between
1844 processes. They copy the values they close over,
1845 but they also own them: that is, no other code can access
1846 them. Owned closures are used in concurrent code, particularly
1847 for spawning [tasks][tasks].
1849 Closures can be used to spawn tasks.
1850 A practical example of this pattern is found when using the `spawn` function,
1851 which starts a new task.
1854 use std::task::spawn;
1856 // proc is the closure which will be spawned.
1858 println!("I'm a new task")
1862 ## Closure compatibility
1864 Rust closures have a convenient subtyping property: you can pass any kind of
1865 closure (as long as the arguments and return types match) to functions
1866 that expect a `||`. Thus, when writing a higher-order function that
1867 only calls its function argument, and does nothing else with it, you
1868 should almost always declare the type of that argument as `||`. That way,
1869 callers may pass any kind of closure.
1872 fn call_twice(f: ||) { f(); f(); }
1873 let closure = || { "I'm a closure, and it doesn't matter what type I am"; };
1874 fn function() { "I'm a normal function"; }
1875 call_twice(closure);
1876 call_twice(function);
1879 > *Note:* Both the syntax and the semantics will be changing
1880 > in small ways. At the moment they can be unsound in some
1881 > scenarios, particularly with non-copyable types.
1885 Methods are like functions except that they always begin with a special argument,
1887 which has the type of the method's receiver. The
1888 `self` argument is like `this` in C++ and many other languages.
1889 Methods are called with dot notation, as in `my_vec.len()`.
1891 _Implementations_, written with the `impl` keyword, can define
1892 methods on most Rust types, including structs and enums.
1893 As an example, let's define a `draw` method on our `Shape` enum.
1896 # fn draw_circle(p: Point, f: f64) { }
1897 # fn draw_rectangle(p: Point, p: Point) { }
1905 Rectangle(Point, Point)
1911 Circle(p, f) => draw_circle(p, f),
1912 Rectangle(p1, p2) => draw_rectangle(p1, p2)
1917 let s = Circle(Point { x: 1.0, y: 2.0 }, 3.0);
1921 This defines an _implementation_ for `Shape` containing a single
1922 method, `draw`. In most respects the `draw` method is defined
1923 like any other function, except for the name `self`.
1925 The type of `self` is the type on which the method is implemented,
1926 or a pointer thereof. As an argument it is written either `self`,
1927 `&self`, or `~self`.
1928 A caller must in turn have a compatible pointer type to call the method.
1931 # fn draw_circle(p: Point, f: f64) { }
1932 # fn draw_rectangle(p: Point, p: Point) { }
1933 # struct Point { x: f64, y: f64 }
1935 # Circle(Point, f64),
1936 # Rectangle(Point, Point)
1939 fn draw_reference(&self) { /* ... */ }
1940 fn draw_owned(~self) { /* ... */ }
1941 fn draw_value(self) { /* ... */ }
1944 let s = Circle(Point { x: 1.0, y: 2.0 }, 3.0);
1946 (&s).draw_reference();
1951 Methods typically take a reference self type,
1952 so the compiler will go to great lengths to convert a callee
1956 # fn draw_circle(p: Point, f: f64) { }
1957 # fn draw_rectangle(p: Point, p: Point) { }
1958 # struct Point { x: f64, y: f64 }
1960 # Circle(Point, f64),
1961 # Rectangle(Point, Point)
1964 # fn draw_reference(&self) { /* ... */ }
1965 # fn draw_owned(~self) { /* ... */ }
1966 # fn draw_value(self) { /* ... */ }
1968 # let s = Circle(Point { x: 1.0, y: 2.0 }, 3.0);
1969 // As with typical function arguments, owned pointers
1970 // are automatically converted to references
1972 (~s).draw_reference();
1974 // Unlike typical function arguments, the self value will
1975 // automatically be referenced ...
1978 // ... and dereferenced
1979 (& &s).draw_reference();
1981 // ... and dereferenced and borrowed
1982 (&~s).draw_reference();
1985 Implementations may also define standalone (sometimes called "static")
1986 methods. The absence of a `self` parameter distinguishes such methods.
1987 These methods are the preferred way to define constructor functions.
1991 fn area(&self) -> f64 { /* ... */ }
1992 fn new(area: f64) -> Circle { /* ... */ }
1996 To call such a method, just prefix it with the type name and a double colon:
1999 use std::f64::consts::PI;
2000 struct Circle { radius: f64 }
2002 fn new(area: f64) -> Circle { Circle { radius: (area / PI).sqrt() } }
2004 let c = Circle::new(42.5);
2009 Throughout this tutorial, we've been defining functions that act only
2010 on specific data types. With type parameters we can also define
2011 functions whose arguments have generic types, and which can be invoked
2012 with a variety of types. Consider a generic `map` function, which
2013 takes a function `function` and a vector `vector` and returns a new
2014 vector consisting of the result of applying `function` to each element
2018 fn map<T, U>(vector: &[T], function: |v: &T| -> U) -> Vec<U> {
2019 let mut accumulator = Vec::new();
2020 for element in vector.iter() {
2021 accumulator.push(function(element));
2027 When defined with type parameters, as denoted by `<T, U>`, this
2028 function can be applied to any type of vector, as long as the type of
2029 `function`'s argument and the type of the vector's contents agree with
2032 Inside a generic function, the names of the type parameters
2033 (capitalized by convention) stand for opaque types. All you can do
2034 with instances of these types is pass them around: you can't apply any
2035 operations to them or pattern-match on them. Note that instances of
2036 generic types are often passed by pointer. For example, the parameter
2037 `function()` is supplied with a pointer to a value of type `T` and not
2038 a value of type `T` itself. This ensures that the function works with
2039 the broadest set of types possible, since some types are expensive or
2040 illegal to copy and pass by value.
2042 Generic `type`, `struct`, and `enum` declarations follow the same pattern:
2045 extern crate collections;
2046 type Set<T> = collections::HashMap<T, ()>;
2059 These declarations can be instantiated to valid types like `Set<int>`,
2060 `Stack<int>`, and `Option<int>`.
2062 The last type in that example, `Option`, appears frequently in Rust code.
2063 Because Rust does not have null pointers (except in unsafe code), we need
2064 another way to write a function whose result isn't defined on every possible
2065 combination of arguments of the appropriate types. The usual way is to write
2066 a function that returns `Option<T>` instead of `T`.
2069 # struct Point { x: f64, y: f64 }
2070 # enum Shape { Circle(Point, f64), Rectangle(Point, Point) }
2071 fn radius(shape: Shape) -> Option<f64> {
2073 Circle(_, radius) => Some(radius),
2074 Rectangle(..) => None
2079 The Rust compiler compiles generic functions very efficiently by
2080 *monomorphizing* them. *Monomorphization* is a fancy name for a simple
2081 idea: generate a separate copy of each generic function at each call site,
2082 a copy that is specialized to the argument
2083 types and can thus be optimized specifically for them. In this
2084 respect, Rust's generics have similar performance characteristics to
2089 Within a generic function—that is, a function parameterized by a
2090 type parameter, say, `T`—the operations we can do on arguments of
2091 type `T` are quite limited. After all, since we don't know what type
2092 `T` will be instantiated with, we can't safely modify or query values
2093 of type `T`. This is where _traits_ come into play. Traits are Rust's
2094 most powerful tool for writing polymorphic code. Java developers will
2095 see them as similar to Java interfaces, and Haskellers will notice
2096 their similarities to type classes. Rust's traits give us a way to
2097 express *bounded polymorphism*: by limiting the set of possible types
2098 that a type parameter could refer to, they expand the number of
2099 operations we can safely perform on arguments of that type.
2101 As motivation, let us consider copying of values in Rust. The `clone`
2102 method is not defined for values of every type. One reason is
2103 user-defined destructors: copying a value of a type that has a
2104 destructor could result in the destructor running multiple times.
2105 Therefore, values of types that have destructors cannot be copied
2106 unless we explicitly implement `clone` for them.
2108 This complicates handling of generic functions.
2109 If we have a function with a type parameter `T`,
2110 can we copy values of type `T` inside that function?
2112 and if we try to run the following code the compiler will complain.
2115 // This does not compile
2116 fn head_bad<T>(v: &[T]) -> T {
2117 v[0] // error: copying a non-copyable value
2121 However, we can tell the compiler
2122 that the `head` function is only for copyable types.
2123 In Rust, copyable types are those that _implement the `Clone` trait_.
2124 We can then explicitly create a second copy of the value we are returning
2125 by calling the `clone` method:
2129 fn head<T: Clone>(v: &[T]) -> T {
2134 The bounded type parameter `T: Clone` says that `head`
2135 can be called on an argument of type `&[T]` for any `T`,
2136 so long as there is an implementation of the
2137 `Clone` trait for `T`.
2138 When instantiating a generic function,
2139 we can only instantiate it with types
2140 that implement the correct trait,
2141 so we could not apply `head` to a vector whose elements are of some type
2142 that does not implement `Clone`.
2144 While most traits can be defined and implemented by user code,
2145 three traits are automatically derived and implemented
2146 for all applicable types by the compiler,
2147 and may not be overridden:
2149 * `Send` - Sendable types.
2151 unless they contain references.
2153 * `Share` - Types that are *threadsafe*
2154 These are types that are safe to be used across several threads with access to
2155 a `&T` pointer. `MutexArc` is an example of a *sharable* type with internal mutable data.
2157 * `'static` - Non-borrowed types.
2158 These are types that do not contain any data whose lifetime is bound to
2159 a particular stack frame. These are types that do not contain any
2160 references, or types where the only contained references
2161 have the `'static` lifetime. (For more on named lifetimes and their uses,
2162 see the [references and lifetimes guide][lifetimes].)
2164 > *Note:* These built-in traits were referred to as 'kinds' in earlier
2165 > iterations of the language, and often still are.
2167 Additionally, the `Drop` trait is used to define destructors. This
2168 trait provides one method called `drop`, which is automatically
2169 called when a value of the type that implements this trait is
2170 destroyed, either because the value went out of scope or because the
2171 garbage collector reclaimed it.
2178 impl Drop for TimeBomb {
2179 fn drop(&mut self) {
2180 for _ in range(0, self.explosivity) {
2187 It is illegal to call `drop` directly. Only code inserted by the compiler
2190 ## Declaring and implementing traits
2192 At its simplest, a trait is a set of zero or more _method signatures_.
2193 For example, we could declare the trait
2194 `Printable` for things that can be printed to the console,
2195 with a single method signature:
2203 We say that the `Printable` trait _provides_ a `print` method with the
2204 given signature. This means that we can call `print` on an argument
2205 of any type that implements the `Printable` trait.
2207 Rust's built-in `Send` and `Share` types are examples of traits that
2208 don't provide any methods.
2210 Traits may be implemented for specific types with [impls]. An impl for
2211 a particular trait gives an implementation of the methods that
2212 trait provides. For instance, the following impls of
2213 `Printable` for `int` and `~str` give implementations of the `print`
2219 # trait Printable { fn print(&self); }
2220 impl Printable for int {
2221 fn print(&self) { println!("{:?}", *self) }
2224 impl Printable for ~str {
2225 fn print(&self) { println!("{}", *self) }
2229 # ("foo".to_owned()).print();
2232 Methods defined in an impl for a trait may be called just like
2233 any other method, using dot notation, as in `1.print()`.
2235 ## Default method implementations in trait definitions
2237 Sometimes, a method that a trait provides will have the same
2238 implementation for most or all of the types that implement that trait.
2239 For instance, suppose that we wanted `bool`s and `f32`s to be
2240 printable, and that we wanted the implementation of `print` for those
2241 types to be exactly as it is for `int`, above:
2244 # trait Printable { fn print(&self); }
2245 impl Printable for f32 {
2246 fn print(&self) { println!("{:?}", *self) }
2249 impl Printable for bool {
2250 fn print(&self) { println!("{:?}", *self) }
2257 This works fine, but we've now repeated the same definition of `print`
2258 in three places. Instead of doing that, we can simply include the
2259 definition of `print` right in the trait definition, instead of just
2260 giving its signature. That is, we can write the following:
2264 // Default method implementation
2265 fn print(&self) { println!("{:?}", *self) }
2268 impl Printable for int {}
2270 impl Printable for ~str {
2271 fn print(&self) { println!("{}", *self) }
2274 impl Printable for bool {}
2276 impl Printable for f32 {}
2279 # ("foo".to_owned()).print();
2284 Here, the impls of `Printable` for `int`, `bool`, and `f32` don't
2285 need to provide an implementation of `print`, because in the absence
2286 of a specific implementation, Rust just uses the _default method_
2287 provided in the trait definition. Depending on the trait, default
2288 methods can save a great deal of boilerplate code from having to be
2289 written in impls. Of course, individual impls can still override the
2290 default method for `print`, as is being done above in the impl for
2293 ## Type-parameterized traits
2295 Traits may be parameterized by type variables. For example, a trait
2296 for generalized sequence types might look like the following:
2300 fn length(&self) -> uint;
2303 impl<T> Seq<T> for ~[T] {
2304 fn length(&self) -> uint { self.len() }
2308 The implementation has to explicitly declare the type parameter that
2309 it binds, `T`, before using it to specify its trait type. Rust
2310 requires this declaration because the `impl` could also, for example,
2311 specify an implementation of `Seq<int>`. The trait type (appearing
2312 between `impl` and `for`) *refers* to a type, rather than
2315 The type parameters bound by a trait are in scope in each of the
2316 method declarations. So, re-declaring the type parameter
2317 `T` as an explicit type parameter for `length`, in either the trait or
2318 the impl, would be a compile-time error.
2320 Within a trait definition, `Self` is a special type that you can think
2321 of as a type parameter. An implementation of the trait for any given
2322 type `T` replaces the `Self` type parameter with `T`. The following
2323 trait describes types that support an equality operation:
2326 // In a trait, `self` refers to the self argument.
2327 // `Self` refers to the type implementing the trait.
2329 fn equals(&self, other: &Self) -> bool;
2332 // In an impl, `self` refers just to the value of the receiver
2334 fn equals(&self, other: &int) -> bool { *other == *self }
2338 Notice that in the trait definition, `equals` takes a
2339 second parameter of type `Self`.
2340 In contrast, in the `impl`, `equals` takes a second parameter of
2341 type `int`, only using `self` as the name of the receiver.
2343 Just as in type implementations, traits can define standalone (static)
2344 methods. These methods are called by prefixing the method name with the trait
2345 name and a double colon. The compiler uses type inference to decide which
2346 implementation to use.
2349 use std::f64::consts::PI;
2350 trait Shape { fn new(area: f64) -> Self; }
2351 struct Circle { radius: f64 }
2352 struct Square { length: f64 }
2354 impl Shape for Circle {
2355 fn new(area: f64) -> Circle { Circle { radius: (area / PI).sqrt() } }
2357 impl Shape for Square {
2358 fn new(area: f64) -> Square { Square { length: area.sqrt() } }
2362 let c: Circle = Shape::new(area);
2363 let s: Square = Shape::new(area);
2366 ## Bounded type parameters and static method dispatch
2368 Traits give us a language for defining predicates on types, or
2369 abstract properties that types can have. We can use this language to
2370 define _bounds_ on type parameters, so that we can then operate on
2374 # trait Printable { fn print(&self); }
2375 fn print_all<T: Printable>(printable_things: ~[T]) {
2376 for thing in printable_things.iter() {
2382 Declaring `T` as conforming to the `Printable` trait (as we earlier
2383 did with `Clone`) makes it possible to call methods from that trait
2384 on values of type `T` inside the function. It will also cause a
2385 compile-time error when anyone tries to call `print_all` on an array
2386 whose element type does not have a `Printable` implementation.
2388 Type parameters can have multiple bounds by separating them with `+`,
2389 as in this version of `print_all` that copies elements.
2392 # trait Printable { fn print(&self); }
2393 fn print_all<T: Printable + Clone>(printable_things: ~[T]) {
2395 while i < printable_things.len() {
2396 let copy_of_thing = printable_things[i].clone();
2397 copy_of_thing.print();
2403 Method calls to bounded type parameters are _statically dispatched_,
2404 imposing no more overhead than normal function invocation, so are
2405 the preferred way to use traits polymorphically.
2407 This usage of traits is similar to Haskell type classes.
2409 ## Trait objects and dynamic method dispatch
2411 The above allows us to define functions that polymorphically act on
2412 values of a single unknown type that conforms to a given trait.
2413 However, consider this function:
2416 # type Circle = int; type Rectangle = int;
2417 # impl Drawable for int { fn draw(&self) {} }
2418 # fn new_circle() -> int { 1 }
2419 trait Drawable { fn draw(&self); }
2421 fn draw_all<T: Drawable>(shapes: ~[T]) {
2422 for shape in shapes.iter() { shape.draw(); }
2424 # let c: Circle = new_circle();
2428 You can call that on an array of circles, or an array of rectangles
2429 (assuming those have suitable `Drawable` traits defined), but not on
2430 an array containing both circles and rectangles. When such behavior is
2431 needed, a trait name can alternately be used as a type, called
2435 # trait Drawable { fn draw(&self); }
2436 fn draw_all(shapes: &[~Drawable]) {
2437 for shape in shapes.iter() { shape.draw(); }
2441 In this example, there is no type parameter. Instead, the `~Drawable`
2442 type denotes any owned box value that implements the `Drawable` trait.
2443 To construct such a value, you use the `as` operator to cast a value
2447 # type Circle = int; type Rectangle = bool;
2448 # trait Drawable { fn draw(&self); }
2449 # fn new_circle() -> Circle { 1 }
2450 # fn new_rectangle() -> Rectangle { true }
2451 # fn draw_all(shapes: &[~Drawable]) {}
2453 impl Drawable for Circle { fn draw(&self) { /* ... */ } }
2454 impl Drawable for Rectangle { fn draw(&self) { /* ... */ } }
2456 let c: ~Circle = ~new_circle();
2457 let r: ~Rectangle = ~new_rectangle();
2458 draw_all([c as ~Drawable, r as ~Drawable]);
2461 We omit the code for `new_circle` and `new_rectangle`; imagine that
2462 these just return `Circle`s and `Rectangle`s with a default size. Note
2463 that, like strings and vectors, objects have dynamic size and may
2464 only be referred to via one of the pointer types.
2465 Other pointer types work as well.
2466 Casts to traits may only be done with compatible pointers so,
2467 for example, an `&Circle` may not be cast to an `~Drawable`.
2470 # type Circle = int; type Rectangle = int;
2471 # trait Drawable { fn draw(&self); }
2472 # impl Drawable for int { fn draw(&self) {} }
2473 # fn new_circle() -> int { 1 }
2474 # fn new_rectangle() -> int { 2 }
2476 let owny: ~Drawable = ~new_circle() as ~Drawable;
2477 // A borrowed object
2478 let stacky: &Drawable = &new_circle() as &Drawable;
2481 Method calls to trait types are _dynamically dispatched_. Since the
2482 compiler doesn't know specifically which functions to call at compile
2483 time, it uses a lookup table (also known as a vtable or dictionary) to
2484 select the method to call at runtime.
2486 This usage of traits is similar to Java interfaces.
2488 There are some built-in bounds, such as `Send` and `Share`, which are properties
2489 of the components of types. By design, trait objects don't know the exact type
2490 of their contents and so the compiler cannot reason about those properties.
2492 You can instruct the compiler, however, that the contents of a trait object must
2493 acribe to a particular bound with a trailing colon (`:`). These are examples of
2500 fn sendable_foo(f: ~Foo:Send) { /* ... */ }
2501 fn shareable_bar<T: Share>(b: &Bar<T>: Share) { /* ... */ }
2504 When no colon is specified (such as the type `~Foo`), it is inferred that the
2505 value ascribes to no bounds. They must be added manually if any bounds are
2506 necessary for usage.
2508 Builtin kind bounds can also be specified on closure types in the same way (for
2509 example, by writing `fn:Send()`), and the default behaviours are the same as
2510 for traits of the same storage class.
2512 ## Trait inheritance
2514 We can write a trait declaration that _inherits_ from other traits, called _supertraits_.
2515 Types that implement a trait must also implement its supertraits.
2517 we can define a `Circle` trait that inherits from `Shape`.
2520 trait Shape { fn area(&self) -> f64; }
2521 trait Circle : Shape { fn radius(&self) -> f64; }
2524 Now, we can implement `Circle` on a type only if we also implement `Shape`.
2527 use std::f64::consts::PI;
2528 # trait Shape { fn area(&self) -> f64; }
2529 # trait Circle : Shape { fn radius(&self) -> f64; }
2530 # struct Point { x: f64, y: f64 }
2531 # fn square(x: f64) -> f64 { x * x }
2532 struct CircleStruct { center: Point, radius: f64 }
2533 impl Circle for CircleStruct {
2534 fn radius(&self) -> f64 { (self.area() / PI).sqrt() }
2536 impl Shape for CircleStruct {
2537 fn area(&self) -> f64 { PI * square(self.radius) }
2541 Notice that methods of `Circle` can call methods on `Shape`, as our
2542 `radius` implementation calls the `area` method.
2543 This is a silly way to compute the radius of a circle
2544 (since we could just return the `radius` field), but you get the idea.
2546 In type-parameterized functions,
2547 methods of the supertrait may be called on values of subtrait-bound type parameters.
2548 Refering to the previous example of `trait Circle : Shape`:
2551 # trait Shape { fn area(&self) -> f64; }
2552 # trait Circle : Shape { fn radius(&self) -> f64; }
2553 fn radius_times_area<T: Circle>(c: T) -> f64 {
2554 // `c` is both a Circle and a Shape
2555 c.radius() * c.area()
2559 Likewise, supertrait methods may also be called on trait objects.
2562 use std::f64::consts::PI;
2563 # trait Shape { fn area(&self) -> f64; }
2564 # trait Circle : Shape { fn radius(&self) -> f64; }
2565 # struct Point { x: f64, y: f64 }
2566 # struct CircleStruct { center: Point, radius: f64 }
2567 # impl Circle for CircleStruct { fn radius(&self) -> f64 { (self.area() / PI).sqrt() } }
2568 # impl Shape for CircleStruct { fn area(&self) -> f64 { PI * square(self.radius) } }
2570 let concrete = ~CircleStruct{center:Point{x:3.0,y:4.0},radius:5.0};
2571 let mycircle: ~Circle = concrete as ~Circle;
2572 let nonsense = mycircle.radius() * mycircle.area();
2575 > *Note:* Trait inheritance does not actually work with objects yet
2577 ## Deriving implementations for traits
2579 A small number of traits in `std` and `extra` can have implementations
2580 that can be automatically derived. These instances are specified by
2581 placing the `deriving` attribute on a data type declaration. For
2582 example, the following will mean that `Circle` has an implementation
2583 for `Eq` and can be used with the equality operators, and that a value
2584 of type `ABC` can be randomly generated and converted to a string:
2590 struct Circle { radius: f64 }
2592 #[deriving(Rand, Show)]
2593 enum ABC { A, B, C }
2596 // Use the Show trait to print "A, B, C."
2597 println!("{}, {}, {}", A, B, C);
2601 The full list of derivable traits is `Eq`, `TotalEq`, `Ord`,
2602 `TotalOrd`, `Encodable` `Decodable`, `Clone`,
2603 `Hash`, `Rand`, `Default`, `Zero`, `FromPrimitive` and `Show`.
2605 # Crates and the module system
2607 Rust's module system is very powerful, but because of that also somewhat complex.
2608 Nevertheless, this section will try to explain every important aspect of it.
2612 In order to speak about the module system, we first need to define the medium it exists in:
2614 Let's say you've written a program or a library, compiled it, and got the resulting binary.
2615 In Rust, the content of all source code that the compiler directly had to compile in order to end up with
2616 that binary is collectively called a 'crate'.
2618 For example, for a simple hello world program your crate only consists of this code:
2623 println!("Hello world!");
2627 A crate is also the unit of independent compilation in Rust: `rustc` always compiles a single crate at a time,
2628 from which it produces either a library or an executable.
2630 Note that merely using an already compiled library in your code does not make it part of your crate.
2632 ## The module hierarchy
2634 For every crate, all the code in it is arranged in a hierarchy of modules starting with a single
2635 root module. That root module is called the 'crate root'.
2637 All modules in a crate below the crate root are declared with the `mod` keyword:
2640 // This is the crate root
2643 // This is the body of module 'farm' declared in the crate root.
2645 fn chicken() { println!("cluck cluck"); }
2646 fn cow() { println!("mooo"); }
2649 // Body of module 'barn'
2651 fn hay() { println!("..."); }
2656 println!("Hello farm!");
2660 As you can see, your module hierarchy is now three modules deep: There is the crate root, which contains your `main()`
2661 function, and the module `farm`. The module `farm` also contains two functions and a third module `barn`,
2662 which contains a function `hay`.
2664 ## Paths and visibility
2666 We've now defined a nice module hierarchy. But how do we access the items in it from our `main` function?
2667 One way to do it is to simply fully qualifying it:
2671 fn chicken() { println!("cluck cluck"); }
2676 println!("Hello chicken!");
2678 ::farm::chicken(); // Won't compile yet, see further down
2682 The `::farm::chicken` construct is what we call a 'path'.
2684 Because it's starting with a `::`, it's also a 'global path', which qualifies
2685 an item by its full path in the module hierarchy relative to the crate root.
2687 If the path were to start with a regular identifier, like `farm::chicken`, it
2688 would be a 'local path' instead. We'll get to them later.
2690 Now, if you actually tried to compile this code example, you'll notice that you
2691 get a `function 'chicken' is private` error. That's because by default, items
2692 (`fn`, `struct`, `static`, `mod`, ...) are private.
2694 To make them visible outside their containing modules, you need to mark them
2695 _public_ with `pub`:
2699 pub fn chicken() { println!("cluck cluck"); }
2700 pub fn cow() { println!("mooo"); }
2705 println!("Hello chicken!");
2706 ::farm::chicken(); // This compiles now
2710 Visibility restrictions in Rust exist only at module boundaries. This
2711 is quite different from most object-oriented languages that also
2712 enforce restrictions on objects themselves. That's not to say that
2713 Rust doesn't support encapsulation: both struct fields and methods can
2714 be private. But this encapsulation is at the module level, not the
2717 Fields are _private_ by default, and can be made _public_ with
2722 # pub type Chicken = int;
2723 # struct Human(int);
2724 # impl Human { pub fn rest(&self) { } }
2725 # pub fn make_me_a_farm() -> Farm { Farm { chickens: ~[], farmer: Human(0) } }
2727 chickens: ~[Chicken],
2732 fn feed_chickens(&self) { /* ... */ }
2733 pub fn add_chicken(&self, c: Chicken) { /* ... */ }
2736 pub fn feed_animals(farm: &Farm) {
2737 farm.feed_chickens();
2742 let f = make_me_a_farm();
2743 f.add_chicken(make_me_a_chicken());
2744 farm::feed_animals(&f);
2747 // This wouldn't compile because both are private:
2748 // `f.feed_chickens();`
2749 // `let chicken_counter = f.chickens.len();`
2751 # fn make_me_a_farm() -> farm::Farm { farm::make_me_a_farm() }
2752 # fn make_me_a_chicken() -> farm::Chicken { 0 }
2755 Exact details and specifications about visibility rules can be found in the Rust
2758 ## Files and modules
2760 One important aspect of Rust's module system is that source files and modules are not the same thing. You define a module hierarchy, populate it with all your definitions, define visibility, maybe put in a `fn main()`, and that's it.
2762 The only file that's relevant when compiling is the one that contains the body
2763 of your crate root, and it's only relevant because you have to pass that file
2764 to `rustc` to compile your crate.
2766 In principle, that's all you need: You can write any Rust program as one giant source file that contains your
2767 crate root and everything else in `mod ... { ... }` declarations.
2769 However, in practice you usually want to split up your code into multiple
2770 source files to make it more manageable. Rust allows you to move the body of
2771 any module into its own source file. If you declare a module without its body,
2772 like `mod foo;`, the compiler will look for the files `foo.rs` and `foo/mod.rs`
2773 inside some directory (usually the same as of the source file containing the
2774 `mod foo;` declaration). If it finds either, it uses the content of that file
2775 as the body of the module. If it finds both, that's a compile error.
2777 To move the content of `mod farm` into its own file, you can write:
2780 // `main.rs` - contains body of the crate root
2781 mod farm; // Compiler will look for `farm.rs` and `farm/mod.rs`
2784 println!("Hello farm!");
2790 // `farm.rs` - contains body of module 'farm' in the crate root
2791 pub fn chicken() { println!("cluck cluck"); }
2792 pub fn cow() { println!("mooo"); }
2795 pub fn hay() { println!("..."); }
2800 In short, `mod foo;` is just syntactic sugar for `mod foo { /* content of <...>/foo.rs or <...>/foo/mod.rs */ }`.
2802 This also means that having two or more identical `mod foo;` declarations
2803 somewhere in your crate hierarchy is generally a bad idea,
2804 just like copy-and-paste-ing a module into multiple places is a bad idea.
2805 Both will result in duplicate and mutually incompatible definitions.
2807 When `rustc` resolves these module declarations, it starts by looking in the
2808 parent directory of the file containing the `mod foo` declaration. For example,
2809 given a file with the module body:
2822 The compiler will look for these files, in this order:
2829 src/animals/fish/mod.rs
2831 src/animals/mammals/humans.rs
2832 src/animals/mammals/humans/mod.rs
2835 Keep in mind that identical module hierarchies can still lead to different path
2836 lookups depending on how and where you've moved a module body to its own file.
2837 For example, if you move the `animals` module into its own file:
2846 // `src/animals.rs` or `src/animals/mod.rs`
2853 ...then the source files of `mod animals`'s submodules can either be in the same directory as the animals source file or in a subdirectory of its directory. If the animals file is `src/animals.rs`, `rustc` will look for:
2860 src/mammals/humans.rs
2861 src/mammals/humans/mod.rs
2864 If the animals file is `src/animals/mod.rs`, `rustc` will look for:
2869 src/animals/fish/mod.rs
2871 src/animals/mammals/humans.rs
2872 src/animals/mammals/humans/mod.rs
2876 These rules allow you to write small modules consisting of single source files which can live in the same directory as well as large modules which group submodule source files in subdirectories.
2878 If you need to override where `rustc` will look for the file containing a
2879 module's source code, use the `path` compiler directive. For example, to load a
2880 `classified` module from a different file:
2883 #[path="../../area51/alien.rs"]
2887 ## Importing names into the local scope
2889 Always referring to definitions in other modules with their global
2890 path gets old really fast, so Rust has a way to import
2891 them into the local scope of your module: `use`-statements.
2893 They work like this: At the beginning of any module body, `fn` body, or any other block
2894 you can write a list of `use`-statements, consisting of the keyword `use` and a __global path__ to an item
2895 without the `::` prefix. For example, this imports `cow` into the local scope:
2899 # mod farm { pub fn cow() { println!("I'm a hidden ninja cow!") } }
2900 # fn main() { cow() }
2903 The path you give to `use` is per default global, meaning relative to the crate root,
2904 no matter how deep the module hierarchy is, or whether the module body it's written in
2905 is contained in its own file. (Remember: files are irrelevant.)
2907 This is different from other languages, where you often only find a single import construct that combines the semantic
2908 of `mod foo;` and `use`-statements, and which tend to work relative to the source file or use an absolute file path
2909 - Ruby's `require` or C/C++'s `#include` come to mind.
2911 However, it's also possible to import things relative to the module of the `use`-statement:
2912 Adding a `super::` in front of the path will start in the parent module,
2913 while adding a `self::` prefix will start in the current module:
2917 # pub fn some_parent_item(){ println!("...") }
2919 use super::some_parent_item;
2920 use self::some_child_module::some_item;
2921 # pub fn bar() { some_parent_item(); some_item() }
2922 # pub mod some_child_module { pub fn some_item() {} }
2927 Again - relative to the module, not to the file.
2929 Imports are also shadowed by local definitions:
2930 For each name you mention in a module/block, `rust`
2931 will first look at all items that are defined locally,
2932 and only if that results in no match look at items you brought in
2933 scope with corresponding `use` statements.
2936 # // FIXME: Allow unused import in doc test
2939 # mod farm { pub fn cow() { println!("Hidden ninja cow is hidden.") } }
2940 fn cow() { println!("Mooo!") }
2943 cow() // resolves to the locally defined `cow()` function
2947 To make this behavior more obvious, the rule has been made that `use`-statement always need to be written
2948 before any declaration, like in the example above. This is a purely artificial rule introduced
2949 because people always assumed they shadowed each other based on order, despite the fact that all items in rust are
2950 mutually recursive, order independent definitions.
2952 One odd consequence of that rule is that `use` statements also go in front of any `mod` declaration,
2953 even if they refer to things inside them:
2958 pub fn cow() { println!("Moooooo?") }
2964 This is what our `farm` example looks like with `use` statements:
2972 pub fn chicken() { println!("cluck cluck"); }
2973 pub fn cow() { println!("mooo"); }
2976 pub fn hay() { println!("..."); }
2981 println!("Hello farm!");
2983 // Can now refer to those names directly:
2990 And here an example with multiple files:
2993 // `a.rs` - crate root
2996 fn main() { foo(); }
3003 pub fn foo() { bar(); }
3008 pub fn bar() { println!("Baz!"); }
3012 There also exist two short forms for importing multiple names at once:
3014 1. Explicit mention multiple names as the last element of an `use` path:
3017 use farm::{chicken, cow};
3019 # pub fn cow() { println!("Did I already mention how hidden and ninja I am?") }
3020 # pub fn chicken() { println!("I'm Bat-chicken, guardian of the hidden tutorial code.") }
3022 # fn main() { cow(); chicken() }
3025 2. Import everything in a module with a wildcard:
3030 # pub fn cow() { println!("Bat-chicken? What a stupid name!") }
3031 # pub fn chicken() { println!("Says the 'hidden ninja' cow.") }
3033 # fn main() { cow(); chicken() }
3036 > *Note:* This feature of the compiler is currently gated behind the
3037 > `#![feature(globs)]` directive. More about these directives can be found in
3040 However, that's not all. You can also rename an item while you're bringing it into scope:
3043 use egg_layer = farm::chicken;
3044 # mod farm { pub fn chicken() { println!("Laying eggs is fun!") } }
3052 In general, `use` creates a local alias:
3053 An alternate path and a possibly different name to access the same item,
3054 without touching the original, and with both being interchangeable.
3056 ## Reexporting names
3058 It is also possible to reexport items to be accessible under your module.
3060 For that, you write `pub use`:
3064 pub use self::barn::hay;
3066 pub fn chicken() { println!("cluck cluck"); }
3067 pub fn cow() { println!("mooo"); }
3070 pub fn hay() { println!("..."); }
3081 Just like in normal `use` statements, the exported names
3082 merely represent an alias to the same thing and can also be renamed.
3084 The above example also demonstrate what you can use `pub use` for:
3085 The nested `barn` module is private, but the `pub use` allows users
3086 of the module `farm` to access a function from `barn` without needing
3087 to know that `barn` exists.
3089 In other words, you can use it to decouple a public api from its internal implementation.
3093 So far we've only talked about how to define and structure your own crate.
3095 However, most code out there will want to use preexisting libraries,
3096 as there really is no reason to start from scratch each time you start a new project.
3098 In Rust terminology, we need a way to refer to other crates.
3100 For that, Rust offers you the `extern crate` declaration:
3104 // `num` ships with Rust (much like `extra`; more details further down).
3107 // The rational number '1/2':
3108 let one_half = ::num::rational::Ratio::new(1, 2);
3112 A statement of the form `extern crate foo;` will cause `rustc` to search for the crate `foo`,
3113 and if it finds a matching binary it lets you use it from inside your crate.
3115 The effect it has on your module hierarchy mirrors aspects of both `mod` and `use`:
3117 - Like `mod`, it causes `rustc` to actually emit code:
3118 The linkage information the binary needs to use the library `foo`.
3120 - But like `use`, all `extern crate` statements that refer to the same library are interchangeable,
3121 as each one really just presents an alias to an external module (the crate root of the library
3122 you're linking against).
3124 Remember how `use`-statements have to go before local declarations because the latter shadows the former?
3125 Well, `extern crate` statements also have their own rules in that regard:
3126 Both `use` and local declarations can shadow them, so the rule is that `extern crate` has to go in front
3127 of both `use` and local declarations.
3129 Which can result in something like this:
3135 use num::rational::Ratio;
3138 pub fn dog() { println!("woof"); }
3143 let a_third = Ratio::new(1, 3);
3147 It's a bit weird, but it's the result of shadowing rules that have been set that way because
3148 they model most closely what people expect to shadow.
3150 ## Crate metadata and settings
3152 For every crate you can define a number of metadata items, such as link name, version or author.
3153 You can also toggle settings that have crate-global consequences. Both mechanism
3154 work by providing attributes in the crate root.
3156 For example, Rust uniquely identifies crates by their link metadata, which includes
3157 the link name and the version. It also hashes the filename and the symbols in a binary
3158 based on the link metadata, allowing you to use two different versions of the same library in a crate
3161 Therefore, if you plan to compile your crate as a library, you should annotate it with that information:
3166 # #![crate_type = "lib"]
3167 #![crate_id = "farm#2.5"]
3173 You can also specify crate id information in a `extern crate` statement. For
3174 example, these `extern crate` statements would both accept and select the
3179 extern crate farm = "farm#2.5";
3180 extern crate my_farm = "farm";
3183 Other crate settings and metadata include things like enabling/disabling certain errors or warnings,
3184 or setting the crate type (library or executable) explicitly:
3190 // This crate is a library ("bin" is the default)
3191 #![crate_id = "farm#2.5"]
3192 #![crate_type = "lib"]
3194 // Turn on a warning
3195 #[warn(non_camel_case_types)]
3199 ## A minimal example
3201 Now for something that you can actually compile yourself.
3203 We define two crates, and use one of them as a library in the other.
3207 #![crate_id = "world#0.42"]
3209 # mod secret_module_to_make_this_test_run {
3210 pub fn explore() -> &'static str { "world" }
3217 fn main() { println!("hello {}", world::explore()); }
3220 Now compile and run like this (adjust to your platform if necessary):
3223 > rustc --crate-type=lib world.rs # compiles libworld-<HASH>-0.42.so
3224 > rustc main.rs -L . # compiles main
3229 Notice that the library produced contains the version in the file name
3230 as well as an inscrutable string of alphanumerics. As explained in the previous paragraph,
3231 these are both part of Rust's library versioning scheme. The alphanumerics are
3232 a hash representing the crate's id.
3234 ## The standard library and the prelude
3236 While reading the examples in this tutorial, you might have asked yourself where all
3237 those magical predefined items like `range` are coming from.
3239 The truth is, there's nothing magical about them: They are all defined normally
3240 in the `std` library, which is a crate that ships with Rust.
3242 The only magical thing that happens is that `rustc` automatically inserts this line into your crate root:
3248 As well as this line into every module body:
3251 use std::prelude::*;
3254 The role of the `prelude` module is to re-export common definitions from `std`.
3256 This allows you to use common types and functions like `Option<T>` or `range`
3257 without needing to import them. And if you need something from `std` that's not in the prelude,
3258 you just have to import it with an `use` statement.
3260 For example, it re-exports `range` which is defined in `std::iter::range`:
3263 use iter_range = std::iter::range;
3266 // `range` is imported by default
3267 for _ in range(0, 10) {}
3269 // Doesn't hinder you from importing it under a different name yourself
3270 for _ in iter_range(0, 10) {}
3272 // Or from not using the automatic import.
3273 for _ in ::std::iter::range(0, 10) {}
3277 Both auto-insertions can be disabled with an attribute if necessary:
3280 // In the crate root:
3286 #![no_implicit_prelude]
3289 See the [API documentation][stddoc] for details.
3291 [stddoc]: std/index.html
3295 Now that you know the essentials, check out any of the additional
3296 guides on individual topics.
3298 * [Pointers][pointers]
3299 * [Lifetimes][lifetimes]
3300 * [Tasks and communication][tasks]
3302 * [The foreign function interface][ffi]
3303 * [Containers and iterators][container]
3304 * [Documenting Rust code][rustdoc]
3305 * [Testing Rust code][testing]
3306 * [The Rust Runtime][runtime]
3308 There is further documentation on the [wiki], however those tend to be even more out of date as this document.
3310 [pointers]: guide-pointers.html
3311 [lifetimes]: guide-lifetimes.html
3312 [tasks]: guide-tasks.html
3313 [macros]: guide-macros.html
3314 [ffi]: guide-ffi.html
3315 [container]: guide-container.html
3316 [testing]: guide-testing.html
3317 [runtime]: guide-runtime.html
3318 [rustdoc]: rustdoc.html
3319 [wiki]: https://github.com/mozilla/rust/wiki/Docs
3321 [wiki-packages]: https://github.com/mozilla/rust/wiki/Doc-packages,-editors,-and-other-tools