Rust Reference Manual
Rust Reference Manual
Contents
1
Introduction
This document is the reference manual for the Rust programming language. It
provides three kinds of material:
Chapters that formally define the language grammar and, for each construct,
informally describe its semantics and give examples of its use.
Chapters that informally describe the memory model, concurrency model,
runtime services, linkage model and debugging facilities.
Appendix chapters providing rationale and references to languages that
influenced the design.
This document does not serve as a tutorial introduction to the language. Background familiarity with the language is assumed. A separate tutorial document
is available to help acquire such background familiarity.
This document also does not serve as a reference to the standard library included
in the language distribution. Those libraries are documented separately by
extracting documentation attributes from their source code.
1.1
Disclaimer
Rust is a work in progress. The language continues to evolve as the design shifts
and is fleshed out in working code. Certain parts work, certain parts do not,
certain parts will be removed or changed.
This manual is a snapshot written in the present tense. All features described
exist in working code unless otherwise noted, but some are quite primitive or
remain to be further modified by planned work. Some may be temporary. It is
a draft, and we ask that you not take anything you read here as final.
If you have suggestions to make, please try to focus them on reductions to the
language: possible features that can be combined or omitted. We aim to keep
the size and complexity of the language under control.
Note: The grammar for Rust given in this document is rough and
very incomplete; only a modest number of sections have accompanying
grammar rules. Formalizing the grammar accepted by the Rust parser
is ongoing work, but future versions of this document will contain a
complete grammar. Moreover, we hope that this grammar will be
extracted and verified as LL(1) by an automated grammar-analysis
tool, and further tested against the Rust sources. Preliminary versions
of this automation exist, but are not yet complete.
Notation
2.1
Unicode productions
2.2
Some rules in the grammar notably unary operators, binary operators, and
keywords are given in a simplified form: as a listing of a table of unquoted,
printable whitespace-separated strings. These cases form a subset of the rules
regarding the token rule, and are assumed to be the result of a lexical-analysis
phase feeding the parser, driven by a DFA, operating over the disjunction of all
such string table entries.
When such a string enclosed in double-quotes (") occurs inside the grammar,
it is an implicit reference to a single member of such a string table production.
See tokens for more information.
3
3.1
Lexical structure
Input format
3.2
The following productions in the Rust grammar are defined in terms of Unicode properties: ident, non_null, non_star, non_eol, non_slash_or_star,
non_single_quote and non_double_quote.
3.2.1
Identifiers
The ident production is any nonempty Unicode string of the following form:
The first character has property XID_start
The remaining characters have property XID_continue
that does not occur in the set of keywords.
Note: XID_start and XID_continue as character properties cover the character
ranges used to form the more familiar C and Java language-family identifiers.
3.2.2
Delimiter-restricted productions
definitions for the special Unicode productions are provided to the grammar
verifier, restricted to ASCII range, when verifying the grammar in this document.
3.3
Comments
3.4
Whitespace
3.5
Tokens
3.5.1
Keywords
Literals
common_escape : \x5c
| n | r | t | 0
| x hex_digit 2
unicode_escape : u hex_digit 4
| U hex_digit 8 ;
hex_digit : a | b | c | d | e | f
| A | B | C | D | E | F
| dec_digit ;
oct_digit : 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 ;
dec_digit : 0 | nonzero_dec ;
nonzero_dec: 1 | 2 | 3 | 4
| 5 | 6 | 7 | 8 | 9 ;
A character literal is a single Unicode character enclosed within two U+0027
(single-quote) characters, with the exception of U+0027 itself, which must be
escaped by a preceding U+005C character (\).
A string literal is a sequence of any Unicode characters enclosed within two
U+0022 (double-quote) characters, with the exception of U+0022 itself, which
must be escaped by a preceding U+005C character (\), or a raw string literal.
Some additional escapes are available in either character or non-raw string literals.
An escape starts with a U+005C (\) and continues with one of the following
forms:
An 8-bit codepoint escape escape starts with U+0078 (x) and is followed
by exactly two hex digits. It denotes the Unicode codepoint equal to the
provided hex value.
A 16-bit codepoint escape starts with U+0075 (u) and is followed by exactly
four hex digits. It denotes the Unicode codepoint equal to the provided
hex value.
A 32-bit codepoint escape starts with U+0055 (U) and is followed by exactly
eight hex digits. It denotes the Unicode codepoint equal to the provided
hex value.
A whitespace escape is one of the characters U+006E (n), U+0072 (r), or
U+0074 (t), denoting the unicode values U+000A (LF), U+000D (CR) or
U+0009 (HT) respectively.
The backslash escape is the character U+005C (\) which must be escaped
in order to denote itself.
Raw string literals do not process any escapes. They start with the character
U+0072 (r), followed by zero or more of the character U+0023 (#) and a U+0022
(double-quote) character. The raw string body is not defined in the EBNF
7
// foo
// "foo"
// R
// \x52
An byte escape escape starts with U+0078 (x) and is followed by exactly
two hex digits. It denotes the byte equal to the provided hex value.
A whitespace escape is one of the characters U+006E (n), U+0072 (r), or
U+0074 (t), denoting the bytes values 0x0A (ASCII LF), 0x0D (ASCII CR)
or 0x09 (ASCII HT) respectively.
The backslash escape is the character U+005C (\) which must be escaped
in order to denote its ASCII encoding 0x5C.
Raw byte string literals do not process any escapes. They start with the character
U+0072 (r), followed by U+0062 (b), followed by zero or more of the character
U+0023 (#), and a U+0022 (double-quote) character. The raw string body is not
defined in the EBNF grammar above: it can contain any sequence of ASCII
characters and is terminated only by another U+0022 (double-quote) character,
followed by the same number of U+0023 (#) characters that preceded the opening
U+0022 (double-quote) character. A raw byte string literal can not contain any
non-ASCII byte.
All characters contained in the raw string body represent their ASCII encoding,
the characters U+0022 (double-quote) (except when followed by at least as many
U+0023 (#) characters as were used to start the raw string literal) or U+005C (\)
do not have any special meaning.
Examples for byte string literals:
b"foo"; br"foo";
b"\"foo\""; br#""foo""#;
// foo
// "foo"
// R
// \x52
Number literals
num_lit : nonzero_dec [
| 0 [
[
| b
[
| o
[
| x
[
dec_digit
dec_digit
1 | 0
oct_digit
hex_digit
|
|
|
|
|
_
_
_
_
_
]
]
]
]
]
int_suffix : u int_suffix_size ?
9
*
*
+
+
+
num_suffix
num_suffix
int_suffix
int_suffix
int_suffix
?
?
?
?
? ] ;
| i int_suffix_size ? ;
int_suffix_size : [ 8 | 1 6 | 3 2 | 6 4 ] ;
float_suffix : [ exponent |
float_suffix_ty : f [ 3
exponent : [E | e] [-
dec_lit : [ dec_digit | _
A decimal literal starts with a decimal digit and continues with any mixture
of decimal digits and underscores.
A hex literal starts with the character sequence U+0030 U+0078 (0x) and
continues as any mixture hex digits and underscores.
An octal literal starts with the character sequence U+0030 U+006F (0o)
and continues as any mixture octal digits and underscores.
A binary literal starts with the character sequence U+0030 U+0062 (0b)
and continues as any mixture binary digits and underscores.
An integer literal may be followed (immediately, without any spaces) by an
integer suffix, which changes the type of the literal. There are two kinds of
integer literal suffix:
The i and u suffixes give the literal type int or uint, respectively.
Each of the signed and unsigned machine types u8, i8, u16, i16, u32, i32,
u64 and i64 give the literal the corresponding machine type.
The type of an unsuffixed integer literal is determined by type inference. If
an integer type can be uniquely determined from the surrounding program
context, the unsuffixed integer literal has that type. If the program context
underconstrains the type, the unsuffixed integer literals type is int; if the
program context overconstrains the type, it is considered a static type error.
Examples of integer literals of various forms:
123; 0xff00;
// information
123u;
123_u;
0xff_u8;
0o70_i16;
0b1111_1111_1001_0000_i32;
Floating-point literals
//
//
//
//
//
type
type
type
type
type
uint
uint
u8
i16
i32
//
//
//
//
type
type
type
type
f64
f64
f32
f64
Unit and boolean literals The unit value, the only value of the type that
has the same name, is written as (). The two values of the boolean type are
written true and false.
3.5.3
Symbols
11
3.6
Paths
12
Syntax extensions
A number of minor features of Rust are not central enough to have their own
syntax, and yet are not implementable as functions. Instead, they are given
names, and invoked through a consistent syntax: name!(...). Examples include:
format! : format data into a string
env! : look up an environment variables value at compile time
file!: return the path to the file being compiled
stringify! : pretty-print the Rust expression given as an argument
include! : include the Rust expression in the given file
include_str! : include the contents of the given file as a string
include_bin! : include the contents of the given file as a binary blob
error!, warn!, info!, debug! : provide diagnostic information.
All of the above extensions are expressions with values.
13
4.1
Macros
Macro By Example
The macro expander matches and transcribes every token that does not begin
with a $ literally, including delimiters. For parsing reasons, delimiters must be
balanced, but they are otherwise not special.
In the matcher, $ name : designator matches the nonterminal in the Rust syntax
named by designator. Valid designators are item, block, stmt, pat, expr, ty
(type), ident, path, matchers (lhs of the => in macro rules), tt (rhs of the =>
in macro rules). In the transcriber, the designator is already known, and so only
the name of a matched nonterminal comes after the dollar sign.
In both the matcher and transcriber, the Kleene star-like operator indicates
repetition. The Kleene star operator consists of $ and parens, optionally followed
by a separator token, followed by * or +. * means zero or more repetitions, +
means at least one repetition. The parens are not matched or transcribed. On
the matcher side, a name is bound to all of the names it matches, in a structure
that mimics the structure of the repetition encountered on a successful match.
The job of the transcriber is to sort that structure out.
The rules for transcription of these repetitions are called Macro By Example. Essentially, one layer of repetition is discharged at a time, and all of
them must be discharged by the time a name is transcribed. Therefore, ( $(
14
Parsing limitations
The parser used by the macro system is reasonably powerful, but the parsing of
Rust syntax is restricted in two ways:
1. The parser will always parse as much as possible. If it attempts to match
$i:expr [ , ] against 8 [ , ], it will attempt to parse i as an array
index operation and fail. Adding a separator can solve this problem.
2. The parser must have eliminated all ambiguity by the time it reaches a $
name : designator. This requirement most often affects name-designator
pairs when they occur at the beginning of, or immediately after, a $(...)*;
requiring a distinctive token in front can solve the problem.
4.2
15
16
Crates contain items, each of which may have some number of attributes attached
to it.
6.1
Items
17
6.1.1
Type Parameters
All items except modules may be parameterized by type. Type parameters are
given as a comma-separated list of identifiers enclosed in angle brackets (<...>),
after the name of the item and before its definition. The type parameters of
an item are considered part of the name, not part of the type of the item.
A referencing path must (in principle) provide type arguments as a list of
comma-separated types enclosed within angle brackets, in order to refer to the
type-parameterized item. In practice, the type-inference system can usually infer
such argument types from context. There are no general type-parametric types,
only type-parametric items. That is, Rust has no notion of type abstraction:
there are no first-class forall types.
6.1.2
Modules
(f64, f64);
-> f64 {
-> f64 {
-> f64 {
18
Modules and types share the same namespace. Declaring a named type that has
the same name as a module in scope is forbidden: that is, a type definition, trait,
struct, enumeration, or type parameter cant shadow the name of a module in
scope, or vice versa.
A module without a body is loaded from an external file, by default with the
same name as the module, plus the .rs extension. When a nested submodule is
loaded from an external file, it is loaded from a subdirectory path that mirrors
the module hierarchy.
// Load the vec module from vec.rs
mod vec;
mod task {
// Load the local_data module from task/local_data.rs
mod local_data;
}
The directories and files used for loading external file modules can be influenced
with the path attribute.
#[path = "task_files"]
mod task {
// Load the local_data module from task_files/tls.rs
#[path = "tls.rs"]
mod local_data;
}
View items
view_item : extern_crate_decl | use_decl ;
A view item manages the namespace of a module. View items do not define new
items, but rather, simply change other items visibility. There are several kinds
of view item:
extern crate declarations
use declarations
Extern crate declarations
extern_crate_decl : "extern" "crate" ident [ ( link_attrs ) ] ? [ = string_lit ] ? ;
link_attrs : link_attr [ , link_attrs ] + ;
link_attr : ident = literal ;
19
Binding all paths matching a given prefix, using the asterisk wildcard
syntax use a::b::*;
An example of use declarations:
use std::iter::range_step;
use std::option::{Some, None};
# fn foo<T>(_: T){}
fn main() {
// Equivalent to std::iter::range_step(0u, 10u, 2u);
range_step(0u, 10u, 2u);
// Equivalent to foo(vec![std::option::Some(1.0f64),
// std::option::None]);
foo(vec![Some(1.0f64), None]);
}
Like items, use declarations are private to the containing module, by default.
Also like items, a use declaration can be public, if qualified by the pub keyword.
Such a use declaration serves to re-export a name. A public use declaration
can therefore redirect some public name to a different target definition: even a
definition with a private canonical path, inside a different module. If a sequence
of such redirections form a cycle or cannot be resolved unambiguously, they
represent a compile-time error.
An example of re-exporting:
# fn main() { }
mod quux {
pub use quux::foo::{bar, baz};
pub mod foo {
pub fn bar() { }
pub fn baz() { }
}
}
In this example, the module quux re-exports two public names defined in foo.
Also note that the paths contained in use items are relative to the crate root.
So, in the previous example, the use refers to quux::foo::{bar, baz}, and not
simply to foo::{bar, baz}. This also means that top-level module declarations
should be at the crate root if direct usage of the declared modules within use
21
items is desired. It is also possible to use self and super at the beginning of
a use item to refer to the current and direct parent modules respectively. All
rules regarding accessing declared modules in use declarations applies to both
module declarations and extern crate declarations.
An example of what will and will not work for use items:
# #![allow(unused_imports)]
use foo::native::start; // good: foo is at the root of the crate
use foo::baz::foobaz;
// good: foo is at the root of the crate
mod foo {
extern crate native;
//
use
use
use
use
foo::native::start;
native::start;
self::baz::foobaz;
foo::bar::foobar;
//
//
//
//
good:
bad:
good:
good:
Functions
An example of a function:
fn add(x: int, y: int) -> int {
return x + y;
}
As with let bindings, function arguments are irrefutable patterns, so any pattern
that is valid in a let binding is also valid as an argument.
fn first((value, _): (int, int)) -> int { value }
Generic functions A generic function allows one or more parameterized types
to appear in its signature. Each type parameter must be explicitly declared, in
an angle-bracket-enclosed, comma-separated list following the function name.
fn iter<T>(seq: &[T], f: |T|) {
for elt in seq.iter() { f(elt); }
}
fn map<T, U>(seq: &[T], f: |T| -> U) -> Vec<U> {
let mut acc = vec![];
for elt in seq.iter() { acc.push(f(elt)); }
acc
}
Inside the function signature and body, the name of the type parameter can be
used as a type name.
When a generic function is referenced, its type is instantiated based on the
context of the reference. For example, calling the iter function defined above
on [1, 2] will instantiate type parameter T with int, and require the closure
parameter to have type fn(int).
The type parameters can also be explicitly supplied in a trailing path component
after the function name. This might be necessary if there is not sufficient context
to determine the type parameters. For example, mem::size_of::<u32>() ==
4.
Since a parameter type is opaque to the generic function, the set of operations
that can be performed on it is limited. Values of parameter type can only be
moved, not copied.
fn id<T>(x: T) -> T { x }
Similarly, trait bounds can be specified for type parameters to allow methods
with that trait to be called on values of that type.
23
Unsafety Unsafe operations are those that potentially violate the memorysafety guarantees of Rusts static semantics.
The following language level features cannot be used in the safe subset of Rust:
Dereferencing a raw pointer.
Reading or writing a mutable static variable.
Calling an unsafe function (including an intrinsic or foreign function).
Unsafe functions Unsafe functions are functions that are not safe in all
contexts and/or for all possible inputs. Such a function must be prefixed with
the keyword unsafe.
Unsafe blocks A block of code can also be prefixed with the unsafe keyword,
to permit calling unsafe functions or dereferencing raw pointers within a safe
function.
When a programmer has sufficient conviction that a sequence of potentially
unsafe operations is actually safe, they can encapsulate that sequence (taken as
a whole) within an unsafe block. The compiler will consider uses of such code
safe, in the surrounding context.
Unsafe blocks are used to wrap foreign libraries, make direct use of hardware
or implement features not directly present in the language. For example, Rust
provides the language features necessary to implement memory-safe concurrency
in the language but the implementation of tasks and message passing is in the
standard library.
Rusts type system is a conservative approximation of the dynamic safety requirements, so in some cases there is a performance cost to using safe code. For
example, a doubly-linked list is not a tree structure and can only be represented
with managed or reference-counted pointers in safe code. By using unsafe blocks
to represent the reverse links as raw pointers, it can be implemented with only
owned pointers.
Behavior considered unsafe This is a list of behavior which is forbidden
in all Rust code. Type checking provides the guarantee that these issues are
never caused by safe code. An unsafe block or function is responsible for never
invoking this behaviour or exposing an API making it possible for it to occur in
safe code.
Data races
Dereferencing a null/dangling raw pointer
24
25
Type definitions
A type definition defines a new name for an existing type. Type definitions are
declared with the keyword type. Every value has a single, specific type; the
type-specified aspects of a value include:
Whether the value is composed of sub-values or is indivisible.
Whether the value represents textual or numerical information.
Whether the value represents integral or floating-point information.
The sequence of memory operations required to access the value.
The kind of the type.
For example, the type (u8, u8) defines the set of immutable values that are
composite pairs, each containing two unsigned 8-bit integers accessed by patternmatching and laid out in memory with the x component preceding the y component.
6.1.5
Structures
27
A tuple structure is a nominal tuple type, also defined with the keyword struct.
For example:
struct Point(int, int);
let p = Point(10, 11);
let px: int = match p { Point(x, _) => x };
A unit-like struct is a structure without any fields, defined by leaving off the list
of fields entirely. Such types will have a single value, just like the unit value ()
of the unit type. For example:
struct Cookie;
let c = [Cookie, Cookie, Cookie, Cookie];
By using the struct_inherit feature gate, structures may use single inheritance.
A Structure may only inherit from a single other structure, called the super-struct.
The inheriting structure (sub-struct) acts as if all fields in the super-struct were
present in the sub-struct. Fields declared in a sub-struct must not have the same
name as any field in any (transitive) super-struct. All fields (both declared and
inherited) must be specified in any initializers. Inheritance between structures
does not give subtyping or coercion. The super-struct and sub-struct must be
defined in the same crate. The super-struct must be declared using the virtual
keyword. For example:
virtual struct Sup { x: int }
struct Sub : Sup { y: int }
let s = Sub {x: 10, y: 11};
let sx = s.x;
6.1.6
Enumerations
Static items
Mutable statics If a static item is declared with the mut keyword, then it
is allowed to be modified by the program. One of Rusts goals is to make
concurrency bugs hard to run into, and this is obviously a very large source
of race conditions or other bugs. For this reason, an unsafe block is required
when either reading or writing a mutable static variable. Care should be taken
to ensure that modifications to a mutable static are safe with respect to other
tasks running in the same process.
Mutable statics are still very useful, however. They can be used with C libraries
and can also be bound from C libraries (in an extern block).
# fn atomic_add(_: &mut uint, _: uint) -> uint { 2 }
static mut LEVELS: uint = 0;
// This violates the idea of no shared state, and this doesnt internally
// protect against races, so this function is unsafe
unsafe fn bump_levels_unsafe1() -> uint {
let ret = LEVELS;
LEVELS += 1;
return ret;
}
// Assuming that we have an atomic_add function which returns the old value,
// this function is "safe" but the meaning of the return value may not be what
// callers expect, so its still marked as unsafe
unsafe fn bump_levels_unsafe2() -> uint {
return atomic_add(&mut LEVELS, 1);
}
6.1.8
Traits
This defines a trait with two methods. All values that have implementations of
this trait in scope can have their draw and bounding_box methods called, using
value.bounding_box() syntax.
Type parameters can be specified for a trait to make it generic. These appear
after the trait name, using the same syntax used in generic functions.
trait
fn
fn
fn
}
Seq<T> {
len(&self) -> uint;
elt_at(&self, n: uint) -> T;
iter(&self, |T|);
Generic functions may use traits as bounds on their type parameters. This will
have two effects: only types that have the trait may instantiate the parameter,
and within the generic function, the methods of the trait can be called on values
that have the parameters type. For example:
# type Surface = int;
# trait Shape { fn draw(&self, Surface); }
fn draw_twice<T: Shape>(surface: Surface, sh: T) {
sh.draw(surface);
sh.draw(surface);
}
Traits also define an object type with the same name as the trait. Values of
this type are created by casting pointer values (pointing to a type for which an
implementation of the given trait is in scope) to pointers to the trait name, used
as a type.
# trait Shape { }
# impl Shape for int { }
# let mycircle = 0i;
let myshape: Box<Shape> = box mycircle as Box<Shape>;
The resulting value is a box containing the value that was cast, along with
information that identifies the methods of the implementation that was used.
Values with a trait type can have methods called on them, for any method in
the trait, and can be used to instantiate type parameters that are bounded by
the trait.
Trait methods may be static, which means that they lack a self argument. This
means that they can only be called with function call syntax (f(x)) and not
method call syntax (obj.f()). The way to refer to the name of a static method
is to qualify it with the trait name, treating the trait name like a module. For
example:
31
trait Num {
fn from_int(n: int) -> Self;
}
impl Num for f64 {
fn from_int(n: int) -> f64 { n as f64 }
}
let x: f64 = Num::from_int(42);
Traits may inherit from other traits. For example, in
trait Shape { fn area() -> f64; }
trait Circle : Shape { fn radius() -> f64; }
the syntax Circle : Shape means that types that implement Circle must
also have an implementation for Shape. Multiple supertraits are separated
by +, trait Circle : Shape + PartialEq { }. In an implementation of
Circle for a given type T, methods can refer to Shape methods, since the
typechecker checks that any type with an implementation of Circle also has an
implementation of Shape.
In type-parameterized functions, methods of the supertrait may be called on
values of subtrait-bound type parameters. Referring to the previous example of
trait Circle : Shape:
# trait Shape { fn area(&self) -> f64; }
# trait Circle : Shape { fn radius(&self) -> f64; }
fn radius_times_area<T: Circle>(c: T) -> f64 {
// c is both a Circle and a Shape
c.radius() * c.area()
}
Likewise, supertrait methods may also be called on trait objects.
# trait Shape { fn area(&self) -> f64; }
# trait Circle : Shape { fn radius(&self) -> f64; }
# impl Shape for int { fn area(&self) -> f64 { 0.0 } }
# impl Circle for int { fn radius(&self) -> f64 { 0.0 } }
# let mycircle = 0;
let mycircle: Circle = ~mycircle as ~Circle;
let nonsense = mycircle.radius() * mycircle.area();
6.1.9
Implementations
External blocks
33
External blocks form the basis for Rusts foreign function interface. Declarations
in an external block describe symbols in external, non-Rust libraries.
Functions within external blocks are declared in the same way as other Rust
functions, with the exception that they may not have a body and are instead
terminated by a semicolon.
extern crate libc;
use libc::{c_char, FILE};
extern {
fn fopen(filename: *c_char, mode: *c_char) -> *FILE;
}
# fn main() {}
Functions within external blocks may be called by Rust code, just like functions
defined in Rust. The Rust compiler automatically translates between the Rust
ABI and the foreign ABI.
A number of attributes control the behavior of external blocks.
By default external blocks assume that the library they are calling uses the
standard C cdecl ABI. Other ABIs may be specified using an abi string, as
shown here:
// Interface to the Windows API
extern "stdcall" { }
The link attribute allows the name of the library to be specified. When specified
the compiler will attempt to link against the native library of the specified name.
#[link(name = "crypto")]
extern { }
The type of a function declared in an extern block is extern "abi" fn(A1,
..., An) -> R, where A1...An are the declared types of its arguments and R
is the declared return type.
6.2
These two terms are often used interchangeably, and what they are attempting
to convey is the answer to the question Can this item be used at this location?
Rusts name resolution operates on a global hierarchy of namespaces. Each
level in the hierarchy can be thought of as some item. The items are one of
34
those mentioned above, but also include external crates. Declaring or defining a
new module can be thought of as inserting a new tree into the hierarchy at the
location of the definition.
To control whether interfaces can be used across modules, Rust checks each use
of an item to see whether it should be allowed or not. This is where privacy
warnings are generated, or otherwise you used a private item of another module
and werent allowed to.
By default, everything in rust is private, with one exception. Enum variants
in a pub enum are also public by default. You are allowed to alter this default
visibility with the priv keyword. When an item is declared as pub, it can be
thought of as being accessible to the outside world. For example:
# fn main() {}
// Declare a private struct
struct Foo;
// Declare a public struct with a private field
pub struct Bar {
field: int
}
// Declare a public enum with two public variants
pub enum State {
PubliclyAccessibleState,
PubliclyAccessibleState2,
}
With the notion of an item being either public or private, Rust allows item
accesses in two cases:
1. If an item is public, then it can be used externally through any of its public
ancestors.
2. If an item is private, it may be accessed by the current module and its
descendants.
These two cases are surprisingly powerful for creating module hierarchies exposing
public APIs while hiding internal implementation details. To help explain, heres
a few use cases and what they would entail.
A library developer needs to expose functionality to crates which link
against their library. As a consequence of the first case, this means that
anything which is usable externally must be pub from the root down to
the destination item. Any private item in the chain will disallow external
accesses.
35
}
// This function is "public to the root" meaning that its available to external
// crates linking against this one.
pub fn public_api() {}
// Similarly to public_api, this module is public so external crates may look
// inside of it.
pub mod submodule {
use crate_helper_module;
36
pub fn my_method() {
// Any item in the local crate may invoke the helper modules public
// interface through a combination of the two rules above.
crate_helper_module::crate_helper();
}
// This function is hidden to any module which is not a descendant of
// submodule
fn my_implementation() {}
#[cfg(test)]
mod test {
#[test]
fn test_my_implementation() {
// Because this module is a descendant of submodule, its allowed
// to access private items inside of submodule without a privacy
// violation.
super::my_implementation();
}
}
}
# fn main() {}
For a rust program to pass the privacy checking pass, all paths must be valid
accesses given the two rules above. This includes all use statements, expressions,
types, etc.
6.2.1
Rust allows publicly re-exporting items through a pub use directive. Because
this is a public directive, this allows the item to be used in the current module
through the rules above. It essentially allows public access into the re-exported
item. For example, this program is valid:
pub use api = self::implementation;
mod implementation {
pub fn f() {}
}
# fn main() {}
37
This means that any external crate referencing implementation::f would receive
a privacy violation, while the path api::f would be allowed.
When re-exporting a private item, it can be thought of as allowing the privacy
chain being short-circuited through the reexport instead of passing through the
namespace hierarchy as it normally would.
6.2.2
6.3
Attributes
attribute : # ! ? [ meta_item ] ;
meta_item : ident [ = literal
| ( meta_seq ) ] ? ;
meta_seq : meta_item [ , meta_seq ] ? ;
Static entities in Rust crates, modules and items may have attributes
applied to them. Attributes in Rust are modeled on Attributes in ECMA-335,
with the syntax coming from ECMA-334 (C#). An attribute is a general, freeform metadatum that is interpreted according to name, convention, and language
and compiler version. Attributes may appear as any of:
A single identifier, the attribute name
An identifier followed by the equals sign = and a literal, providing a
key/value pair
An identifier followed by a parenthesized list of sub-attribute arguments
Attributes with a bang (!) after the hash (#) apply to the item that the
attribute is declared within. Attributes that do not have a bang after the hash
apply to the item that follows the attribute.
An example of attributes:
38
Crate-only attributes
Module-only attributes
macro_escape - macros defined in this module will be visible in the modules parent, after this module has been included.
39
Function-only attributes
plugin_registrar - mark this function as the registration point for compiler plugins, such as loadable syntax extensions.
main - indicates that this function should be passed to the entry point,
rather than the function in the crate root named main.
start - indicates that this function should be used as the entry point,
overriding the start language item. See the start language item for
more details.
6.3.4
Static-only attributes
thread_local - on a static mut, this signals that the value of this static
may change depending on the current thread. The exact consequences of
this are implementation-defined.
6.3.5
FFI attributes
40
6.3.6
Miscellaneous attributes
Conditional compilation
Sometimes one wants to have different compiler outputs from the same code,
depending on build target, such as targeted operating system, or to enable release
builds.
There are two kinds of configuration options, one that is either defined or not
(#[cfg(foo)]), and the other that contains a string that can be checked against
(#[cfg(bar = "baz")] (currently only compiler-defined configuration options
can have the latter form).
41
// The function is only included in the build when compiling for OSX
#[cfg(target_os = "macos")]
fn macos_only() {
// ...
}
// This function is only included when either foo or bar is defined
#[cfg(foo)]
#[cfg(bar)]
fn needs_foo_or_bar() {
// ...
}
// This function is only included when compiling for a unixish OS with a 32-bit
// architecture
#[cfg(unix, target_word_size = "32")]
fn on_32bit_unix() {
// ...
}
This illustrates some conditional compilation can be achieved using the
#[cfg(...)] attribute. Note that #[cfg(foo, bar)] is a condition that needs
both foo and bar to be defined while #[cfg(foo)] #[cfg(bar)] only needs
one of foo and bar to be defined (this resembles in the disjunctive normal form).
Additionally, one can reverse a condition by enclosing it in a not(...), like e. g.
#[cfg(not(target_os = "win32"))].
The following configurations must be defined by the implementation:
target_arch = "...".
Target CPU architecture, such as "x86",
"x86_64" "mips", or "arm".
target_endian = "...". Endianness of the target CPU, either "little"
or "big".
target_family = "...". Operating system family of the target, e. g.
"unix" or "windows". The value of this configuration option is defined as
a configuration itself, like unix or windows.
target_os = "...". Operating system of the target, examples include
"win32", "macos", "linux", "android" or "freebsd".
target_word_size = "...". Target word size in bits. This is set to "32"
for targets with 32-bit pointers, and likewise set to "64" for 64-bit pointers.
unix. See target_family.
windows. See target_family.
42
6.3.8
#[warn(missing_doc)]
pub fn undocumented_two() -> int { 2 }
}
// Missing documentation signals a warning here
pub fn undocumented_too() -> int { 3 }
}
This example shows how one can use forbid to disallow uses of allow for that
lint check.
#[forbid(missing_doc)]
mod m3 {
// Attempting to toggle warning signals an error here
#[allow(missing_doc)]
/// Returns 2.
pub fn undocumented_too() -> int { 2 }
}
6.3.9
Language items
Some primitive Rust operations are defined in Rust code, rather than being
implemented directly in C or assembly language. The definitions of these
operations have to be easy for the compiler to find. The lang attribute makes it
possible to declare these operations. For example, the str module in the Rust
standard library defines the string equality function:
#[lang="str_eq"]
pub fn eq_slice(a: &str, b: &str) -> bool {
// details elided
}
The name str_eq has a special meaning to the Rust compiler, and the presence
of this definition means that it will use this definition when generating calls to
the string equality function.
A complete list of the built-in language items follows:
Built-in Traits
send Able to be sent across task boundaries.
sized Has a size known at compile time.
copy Types that do not move ownership when used by-value.
44
45
Types
unsafe A type whose contents can be mutated through an immutable
reference
type_id The type returned by the type_id intrinsic.
Marker types
46
6.3.10
Inline attributes
Deriving
Stability
One can indicate the stability of an API using the following attributes:
deprecated: This item should no longer be used, e.g. it has been replaced.
No guarantee of backwards-compatibility.
experimental: This item was only recently introduced or is otherwise
in a state of flux. It may change significantly, or even be removed. No
guarantee of backwards-compatibility.
unstable: This item is still under development, but requires more testing
to be considered stable. No guarantee of backwards-compatibility.
stable: This item is considered stable, and will not change significantly.
Guarantee of backwards-compatibility.
frozen: This item is very stable, and is unlikely to change. Guarantee of
backwards-compatibility.
locked: This item will never change unless a serious bug is found. Guarantee of backwards-compatibility.
48
6.3.13
Compiler Features
Certain aspects of Rust may be implemented in the compiler, but theyre not
necessarily ready for every-day use. These features are often of prototype
quality or almost production ready, but may not be stable enough to be
considered a full-fledged language feature.
For this reason, Rust recognizes a special crate-level attribute of the form:
#![feature(feature1, feature2, feature3)]
This directive informs the compiler that the feature list: feature1, feature2,
and feature3 should all be enabled. This is only recognized at a crate-level, not
at a module-level. Without this directive, all features are considered off, and
using the features will result in a compiler error.
The currently implemented features of the reference compiler are:
macro_rules - The definition of new macros. This does not encompass
macro-invocation, that is always enabled by default, this only covers the
definition of new macros. There are currently various problems with
invoking macros, how they interact with their environment, and possibly
how they are used outside of location in which they are defined. Macro
definitions are likely to change slightly in the future, so they are currently
hidden behind this feature.
globs - Importing everything in a module through *. This is currently a
large source of bugs in name resolution for Rust, and its not clear whether
this will continue as a feature or not. For these reasons, the glob import
statement has been hidden behind this feature flag.
struct_variant - Structural enum variants (those with named fields).
It is currently unknown whether this style of enum variant is as fully
supported as the tuple-forms, and its not certain that this style of variant
should remain in the language. For now this style of variant is hidden
behind a feature flag.
once_fns - Onceness guarantees a closure is only executed once. Defining
a closure as once is unlikely to be supported going forward. So they are
hidden behind this feature until they are to be removed.
managed_boxes - Usage of @ pointers is gated due to many planned changes
to this feature. In the past, this has meant a GC pointer, but the current
implementation uses reference counting and will likely change drastically
over time. Additionally, the @ syntax will no longer be used to create GC
boxes.
50
asm - The asm! macro provides a means for inline assembly. This is often
useful, but the exact syntax for this feature along with its semantics are
likely to change, so this macro usage must be opted into.
non_ascii_idents - The compiler supports the use of non-ascii identifiers,
but the implementation is a little rough around the edges, so this can be
seen as an experimental feature for now until the specification of identifiers
is fully fleshed out.
thread_local - The usage of the #[thread_local] attribute is experimental and should be seen as unstable. This attribute is used to declare
a static as being unique per-thread leveraging LLVMs implementation
which works in concert with the kernel loader and dynamic linker. This is
not necessarily available on all platforms, and usage of it is discouraged
(rust focuses more on task-local data instead of thread-local data).
link_args - This attribute is used to specify custom flags to the linker,
but usage is strongly discouraged. The compilers usage of the system
linker is not guaranteed to continue in the future, and if the system linker
is not used then specifying custom flags doesnt have much meaning.
If a feature is promoted to a language feature, then all existing programs will
start to receive compilation warnings about #[feature] directives which enabled
the new feature (because the directive is no longer necessary). However, if a
feature is decided to be removed from the language, errors will be issued (if there
isnt a parser error first). The directive in this case is no longer necessary, and
its likely that existing code will break if the feature isnt removed.
If a unknown feature is found in a directive, it results in a compiler error. An
unknown feature is one which has never been recognized by the compiler.
51
7.1
Statements
Declaration statements
A declaration statement is one that introduces one or more names into the
enclosing statement block. The declared names may denote new slots or new
items.
Item declarations An item declaration statement has a syntactic form identical to an item declaration within a module. Declaring an item a function,
enumeration, structure, type, static, trait, implementation or module locally
within a statement block is simply a way of restricting its scope to a narrow
region containing all of its uses; it is otherwise identical in meaning to declaring
the item outside the statement block.
Note: there is no implicit capture of the functions dynamic environment when
declaring a function-local item.
Slot declarations
let_decl : "let" pat [: type ] ? [ init ] ? ; ;
init : [ = ] expr ;
A slot declaration introduces a new set of slots, given by a pattern. The pattern
may be followed by a type annotation, and/or an initializer expression. When
no type annotation is given, the compiler will infer the type, or signal an error
if insufficient type information is available for definite inference. Any slots
introduced by a slot declaration are visible from the point of declaration until
the end of the enclosing block scope.
7.1.2
Expression statements
52
7.2
Expressions
An expression may have two roles: it always produces a value, and it may have
effects (otherwise known as side effects). An expression evaluates to a value,
and has effects during evaluation. Many expressions contain sub-expressions
(operands). The meaning of each kind of expression dictates several things: *
Whether or not to evaluate the sub-expressions when evaluating the expression
* The order in which to evaluate the sub-expressions * How to combine the
sub-expressions values to obtain the value of the expression.
In this way, the structure of expressions dictates the structure of execution.
Blocks are just another kind of expression, so blocks, statements, expressions,
and blocks again can recursively nest inside each other to an arbitrary depth.
Lvalues, rvalues and temporaries Expressions are divided into two main
categories: lvalues and rvalues. Likewise within each expression, sub-expressions
may occur in lvalue context or rvalue context. The evaluation of an expression
depends both on its own category and the context it occurs within.
An lvalue is an expression that represents a memory location. These expressions
are paths (which refer to local variables, function and method arguments, or
static variables), dereferences (*expr), indexing expressions (expr[expr]), and
field references (expr.f). All other expressions are rvalues.
The left operand of an assignment or compound-assignment expression is an
lvalue context, as is the single operand of a unary borrow. All other expression
contexts are rvalue contexts.
When an lvalue is evaluated in an lvalue context, it denotes a memory location;
when evaluated in an rvalue context, it denotes the value held in that memory
location.
When an rvalue is used in lvalue context, a temporary un-named lvalue is created
and used instead. A temporarys lifetime equals the largest lifetime of any
reference that points to it.
Moved and copied types When a local variable is used as an rvalue the
variable will either be moved or copied, depending on its type. For types that
contain owning pointers or values that implement the special trait Drop, the
variable is moved. All other types are copied.
7.2.1
Literal expressions
53
();
"hello";
5;
5;
7.2.2
//
//
//
//
unit type
string type
character type
integer type
Path expressions
Tuple expressions
Structure expressions
struct_expr : expr_path
[
[
expr_path
[
expr_path
{ ident : expr
, ident : expr ] *
".." expr ] } |
( expr
, expr ] * ) |
;
54
Block expressions
block_expr : { [ view_item ] *
[ stmt ; | item ] *
[ expr ] } ;
A block expression is similar to a module in terms of the declarations that are
possible. Each block conceptually introduces a new namespace scope. View
items can bring new names into scopes and declared items are in scope for only
the block itself.
A block will execute each statement sequentially, and then execute the expression
(if given). If the final expression is omitted, the type and return value of the
block are (), but if it is provided, the type and return value of the block are
that of the expression itself.
7.2.6
Method-call expressions
Field expressions
Vector expressions
56
7.2.9
Index expressions
Rust defines six symbolic unary operators. They are all written as prefix
operators, before the expression they apply to.
- Negation. May only be applied to numeric types.
* Dereference. When applied to a pointer it denotes the pointed-to location. For pointers to mutable locations, the resulting lvalue can
be assigned to. On non-pointer types, it calls the deref method
of the std::ops::Deref trait, or the deref_mut method of the
std::ops::DerefMut trait (if implemented by the type and required
for an outer expression that will or could mutate the dereference), and
produces the result of dereferencing the & or &mut borrowed pointer
returned from the overload method.
! Logical negation. On the boolean type, this flips between true and
false. On integer types, this inverts the individual bits in the twos
complement representation of the value.
box Boxing operators. Allocate a box to hold the value they are applied
to, and store the value in it. box creates an owned box.
& Borrow operator. Returns a reference, pointing to its operand. The
operand of a borrow is statically proven to outlive the resulting pointer.
If the borrow-checker cannot prove this, it is a compilation error.
57
7.2.11
58
Comparison operators Comparison operators are, like the arithmetic operators, and bitwise operators, syntactic sugar for calls to built-in traits. This
means that comparison operators can be overridden for user-defined types. The
default meaning of the operators on standard types is given here.
== Equal to. Calls the eq method on the std::cmp::PartialEq trait.
!= Unequal to. Calls the ne method on the std::cmp::PartialEq trait.
< Less than. Calls the lt method on the std::cmp::PartialOrd trait.
> Greater than. Calls the gt method on the std::cmp::PartialOrd trait.
<= Less than or equal. Calls the le method on the std::cmp::PartialOrd
trait.
>= Greater than or equal. Calls the ge method on the std::cmp::PartialOrd
trait.
Type cast expressions
operator as.
Executing an as expression casts the value on the left-hand side to the type on
the right-hand side.
A numeric value can be cast to any numeric type. A raw pointer value can
be cast to or from any integral type or raw pointer type. Any other cast is
unsupported and will fail to compile.
An example of an as expression:
# fn sum(v: &[f64]) -> f64 { 0.0 }
# fn len(v: &[f64]) -> int { 0 }
fn avg(v: &[f64]) -> f64 {
let sum: f64 = sum(v);
let sz: f64 = len(v) as f64;
return sum / sz;
}
Assignment expressions An assignment expression consists of an lvalue
expression followed by an equals sign (=) and an rvalue expression.
Evaluating an assignment expression either copies or moves its right-hand operand
to its left-hand operand.
# let mut x = 0;
# let y = 0;
x = y;
59
Grouped expressions
Call expressions
A call expression invokes a function, providing zero or more input slots and
an optional reference slot to serve as the functions output, bound to the lval
on the right hand side of the call. If the function eventually returns, then the
expression completes.
Some examples of call expressions:
# use std::from_str::FromStr;
# fn add(x: int, y: int) -> int { 0 }
let x: int = add(1, 2);
let pi: Option<f32> = FromStr::from_str("3.14");
7.2.14
Lambda expressions
i += 1;
}
}
ten_times(|j| println!("hello, {}", j));
7.2.15
While loops
Infinite loops
Break expressions
7.2.18
Continue expressions
For expressions
63
7.2.20
If expressions
Match expressions
The first pattern matches lists constructed by applying Cons to any head value,
and a tail value of box Nil. The second pattern matches any list constructed
with Cons, ignoring the values of its arguments. The difference between _ and
.. is that the pattern C(_) is only type-correct if C has exactly one argument,
while the pattern C(..) is type-correct for any enum variant C, regardless of
how many arguments C has.
Used inside a vector pattern, .. stands for any number of elements. This
wildcard can be used at most once for a given vector, which implies that it
cannot be used to specifically match elements that are at an unknown distance
from both ends of a vector, like [.., 42, ..]. If followed by a variable name,
it will bind the corresponding slice to the variable. Example:
fn is_symmetric(list: &[uint]) -> bool {
match list {
[] | [_]
=> true,
[x, ..inside, y] if x == y => is_symmetric(inside),
_
=> false
}
}
fn main() {
let sym
= &[0, 1, 4, 2, 4, 1, 0];
let not_sym = &[0, 1, 7, 2, 4, 1, 0];
assert!(is_symmetric(sym));
assert!(!is_symmetric(not_sym));
}
A match behaves differently depending on whether or not the head expression is
an lvalue or an rvalue. If the head expression is an rvalue, it is first evaluated
into a temporary location, and the resulting value is sequentially compared to
the patterns in the arms until a match is found. The first arm with a matching
pattern is chosen as the branch target of the match, any variables bound by the
pattern are assigned to local variables in the arms block, and control enters the
block.
When the head expression is an lvalue, the match does not allocate a temporary
location (however, a by-value binding may copy or move from the lvalue). When
possible, it is preferable to match on lvalues, as the lifetime of these matches
inherits the lifetime of the lvalue, rather than being restricted to the inside of
the match.
An example of a match expression:
# fn process_pair(a: int, b: int) { }
# fn process_ten() { }
65
67
7.2.22
Return expressions
Type system
8.1
Types
Every slot, item and value in a Rust program has a type. The type of a value
defines the interpretation of the memory holding it.
Built-in types and type-constructors are tightly integrated into the language,
in nontrivial ways that are not possible to emulate in user-defined types. Userdefined types have limited capabilities.
8.1.1
Primitive types
68
Machine types
The unsigned word types u8, u16, u32 and u64, with values drawn from
the integer intervals [0, 28 - 1], [0, 216 - 1], [0, 232 - 1] and [0, 264 - 1]
respectively.
The signed twos complement word types i8, i16, i32 and i64, with values
drawn from the integer intervals [-(2(7)), 27 - 1], [-(2(15)), 215 - 1],
[-(2(31)), 231 - 1], [-(2(63)), 263 - 1] respectively.
The IEEE 754-2008 binary32 and binary64 floating-point types: f32 and
f64, respectively.
Machine-dependent integer types The Rust type uint 4 is an unsigned
integer type with target-machine-dependent size. Its size, in bits, is equal to the
number of bits required to hold any memory address on the target machine.
The Rust type int 5 is a twos complement signed integer type with targetmachine-dependent size. Its size, in bits, is equal to the size of the rust type
uint on the same target machine.
8.1.2
Textual types
Tuple types
69
The members of a tuple are laid out in memory contiguously, in order specified
by the tuple type.
An example of a tuple type and its use:
type Pair<a> = (int, &a str);
let p: Pair<static> = (10, "hello");
let (a, b) = p;
assert!(b != "world");
8.1.4
Vector types
Structure types
A struct type is a heterogeneous product of other types, called the fields of the
type.6
New instances of a struct can be constructed with a struct expression.
The memory layout of a struct is undefined by default to allow for compiler
optimziations like field reordering, but it can be fixed with the #[repr(...)]
6 struct types are analogous struct types in C, the record types of the ML family, or the
structure types of the Lisp family.
70
Enumerated types
Recursive types
71
Recursive type definitions can cross module boundaries, but not module
visibility boundaries, or crate boundaries (in order to simplify the module
system and type checker).
An example of a recursive type and its use:
enum List<T> {
Nil,
Cons(T, Box<List<T>>)
}
let a: List<int> = Cons(7, box Cons(13, box Nil));
8.1.8
Pointer types
All pointers in Rust are explicit first-class values. They can be copied, stored
into data structures, and returned from functions. There are four varieties of
pointer in Rust:
Owning pointers (Box) These point to owned heap allocations (or
boxes) in the shared, inter-task heap. Each owned box has a single
owning pointer; pointer and pointee retain a 1:1 relationship at all
times. Owning pointers are written Box<content>, for example
Box<int> means an owning pointer to an owned box containing an
integer. Copying an owned box is a deep operation: it involves
allocating a new owned box and copying the contents of the old box
into the new box. Releasing an owning pointer immediately releases
its corresponding owned box.
References (&) These point to memory owned by some other value. References arise by (automatic) conversion from owning pointers, managed
pointers, or by applying the borrowing operator & to some other value,
including lvalues, rvalues or temporaries. References are written
&content, or in some cases &f content for some lifetime-variable f,
for example &int means a reference to an integer. Copying a reference
is a shallow operation: it involves only copying the pointer itself.
Releasing a reference typically has no effect on the value it points to,
with the exception of temporary values, which are released when the
last reference to them is released.
Raw pointers (*) Raw pointers are pointers without safety or liveness
guarantees. Raw pointers are written as *const T or *mut T, for
example *const int means a raw pointer to an integer. Copying or
dropping a raw pointer has no effect on the lifecycle of any other value.
Dereferencing a raw pointer or converting it to any other pointer type
72
Function types
The function type constructor fn forms new function types. A function type
consists of a possibly-empty set of function-type modifiers (such as unsafe or
extern), a sequence of input types and an output type.
An example of a fn type:
fn add(x: int, y: int) -> int {
return x + y;
}
let mut x = add(5,7);
type Binop<a> = |int,int|: a -> int;
let bo: Binop = add;
x = bo(5,7);
8.1.10
Closure types
Object types
Every trait item (see traits) defines a type with the same name as the trait.
This type is called the object type of the trait. Object types permit late
binding of methods, dispatched using virtual method tables (vtables). Whereas
most calls to trait methods are early bound (statically resolved) to specific
implementations at compile time, a call to a method on an object type is only
resolved to a vtable entry at compile time. The actual implementation for each
vtable entry can vary on an object-by-object basis.
Given a pointer-typed expression E of type &T or Box<T>, where T implements
trait R, casting E to the corresponding pointer type &R or Box<R> results in a
value of the object type R. This result is represented as a pair of pointers: the
vtable pointer for the T implementation of R, and the pointer value of E.
74
Type parameters
Within the body of an item that has type parameter declarations, the names of
its type parameters are types:
fn map<A: Clone, B: Clone>(f: |A| -> B, xs: &[A]) -> Vec<B> {
if xs.len() == 0 {
return vec![];
}
let first: B = f(xs[0].clone());
let rest: Vec<B> = map(f, xs.slice(1, xs.len()));
return vec![first].append(rest.as_slice());
}
Here, first has type B, referring to maps B type parameter; and rest has type
Vec<B>, a vector type with element type B.
8.1.13
Self types
The special type self has a meaning within methods inside an impl item. It
refers to the type of the implicit self argument. For example, in:
75
trait Printable {
fn make_string(&self) -> String;
}
impl Printable for String {
fn make_string(&self) -> String {
(*self).clone()
}
}
self refers to the value of type String that is the receiver for a call to the
method make_string.
8.2
Type kinds
Types in Rust are categorized into kinds, based on various properties of the
components of the type. The kinds are:
Send Types of this kind can be safely sent between tasks. This kind
includes scalars, owning pointers, owned closures, and structural types
containing only other owned types. All Send types are static.
Copy Types of this kind consist of Plain Old Data which can be copied
by simply moving bits. All values of this kind can be implicitly
copied. This kind includes scalars and immutable references, as well
as structural types containing other Copy types.
static Types of this kind do not contain any references (except for
references with the static lifetime, which are allowed). This can be
a useful guarantee for code that breaks borrowing assumptions using
unsafe operations.
Drop This is not strictly a kind, but its presence interacts with kinds: the
Drop trait provides a single method drop that takes no parameters,
and is run when values of the type are dropped. Such a method is
called a destructor, and are always executed in top-down order:
a value is completely destroyed before any of the values it owns run
their destructors. Only Send types can implement Drop.
Default Types with destructors, closure environments, and various other
non-first-class types, are not copyable at all. Such types can usually
only be accessed through pointers, or in some cases, moved between
mutable locations.
Kinds can be supplied as bounds on type parameters, like traits, in which case
the parameter is constrained to types satisfying that kind.
76
By default, type parameters do not carry any assumed kind-bounds at all. When
instantiating a type parameter, the kind bounds on the parameter are checked
to be the same or narrower than the kind of the type that it is instantiated with.
Sending operations are not part of the Rust language, but are implemented in
the library. Generic functions that send values bound the kind of these values to
sendable.
9.1
Memory model
A Rust programs memory consists of a static set of items, a set of tasks each
with its own stack, and a heap. Immutable portions of the heap may be shared
between tasks, mutable portions may not.
Allocations in the stack consist of slots, and allocations in the heap consist of
boxes.
9.1.1
The items of a program are those functions, modules and types that have their
value calculated at compile-time and stored uniquely in the memory image of
the rust process. Items are neither dynamically allocated nor freed.
A tasks stack consists of activation frames automatically allocated on entry to
each function as the task executes. A stack allocation is reclaimed when control
leaves the frame containing it.
The heap is a general term that describes two separate sets of boxes: managed
boxes which may be subject to garbage collection and owned boxes. The
lifetime of an allocation in the heap depends on the lifetime of the box values
pointing to it. Since box values may themselves be passed in and out of frames,
or stored in the heap, heap allocations may outlive the frame they are allocated
within.
77
9.1.2
Memory ownership
A task owns all memory it can safely reach through local variables, as well as
managed, owned boxes and references.
When a task sends a value that has the Send trait to another task, it loses
ownership of the value sent and can no longer refer to it. This is statically
guaranteed by the combined use of move semantics, and the compiler-checked
meaning of the Send trait: it is only instantiated for (transitively) sendable kinds
of data constructor and pointers, never including managed boxes or references.
When a stack frame is exited, its local allocations are all released, and its
references to boxes (both managed and owned) are dropped.
A managed box may (in the case of a recursive, mutable managed type) be
cyclic; in this case the release of memory inside the managed structure may be
deferred until task-local garbage collection can reclaim it. Code can ensure no
such delayed deallocation occurs by restricting itself to owned boxes and similar
unmanaged kinds of data.
When a task finishes, its stack is necessarily empty and it therefore has no
references to any boxes; the remainder of its heap is immediately freed.
9.1.3
Memory slots
78
Local variables are not initialized when allocated; the entire frame worth of
local variables are allocated at once, on frame-entry, in an uninitialized state.
Subsequent statements within a function may or may not initialize the local
variables. Local variables can be used only after they have been initialized; this
is enforced by the compiler.
9.1.4
Owned boxes
9.2
Tasks
79
9.2.1
Rust tasks are isolated and generally unable to interfere with one anothers
memory directly, except through unsafe code. All contact between tasks is
mediated by safe forms of ownership transfer, and data races on memory are
prohibited by the type system.
Inter-task communication and co-ordination facilities are provided in the standard
library. These include:
synchronous and asynchronous communication channels with various communication topologies
read-only and read-write shared variables with various safe mutual exclusion
patterns
simple locks and semaphores
When such facilities carry values, the values are restricted to the Send type-kind.
Restricting communication interfaces to this kind ensures that no references or
managed pointers move between tasks. Thus access to an entire data structure
can be mediated through its owning root value; no further locking or copying
is required to avoid data races within the substructure of such a value.
9.2.2
Task lifecycle
The lifecycle of a task consists of a finite set of states and events that cause
transitions between the states. The lifecycle states of a task are:
running
blocked
failing
dead
A task begins its lifecycle once it has been spawned in the running state.
In this state it executes the statements of its entry function, and any functions
called by the entry function.
A task may transition from the running state to the blocked state any time it
makes a blocking communication call. When the call can be completed when
a message arrives at a sender, or a buffer opens to receive a message then the
blocked task will unblock and transition back to running.
A task may transition to the failing state at any time, due being killed by some
external event or internally, from the evaluation of a fail!() macro. Once
80
failing, a task unwinds its stack and transitions to the dead state. Unwinding
the stack of a task is done by the task itself, on its own control stack. If a
value with a destructor is freed during unwinding, the code for the destructor
is run, also on the tasks control stack. Running the destructor code causes a
temporary transition to a running state, and allows the destructor code to cause
any subsequent state transitions. The original task of unwinding and failing
thereby may suspend temporarily, and may involve (recursive) unwinding of the
stack of a failed destructor. Nonetheless, the outermost unwinding activity will
continue until the stack is unwound and the task transitions to the dead state.
There is no way to recover from task failure. Once a task has temporarily
suspended its unwinding in the failing state, failure occurring from within this
destructor results in hard failure. A hard failure currently results in the process
aborting.
A task in the dead state cannot transition to other states; it exists only to have
its termination status inspected by other tasks, and/or to await reclamation
when the last reference to it drops.
9.2.3
Task scheduling
The currently scheduled task is given a finite time slice in which to execute, after
which it is descheduled at a loop-edge or similar preemption point, and another
task within is scheduled, pseudo-randomly.
An executing task can yield control at any time, by making a library call
to std::task::yield, which deschedules it immediately. Entering any other
non-executing state (blocked, dead) similarly deschedules the task.
10
The Rust runtime is a relatively compact collection of C++ and Rust code
that provides fundamental services and datatypes to all Rust tasks at run-time.
It is smaller and simpler than many modern language runtimes. It is tightly
integrated into the languages execution model of memory, tasks, communication
and logging.
Note: The runtime library will merge with the std library in future
versions of Rust.
10.0.4
Memory allocation
Built in types
The runtime provides C and Rust code to assist with various built-in types, such
as vectors, strings, and the low level communication system (ports, channels,
tasks).
Support for other built-in types such as simple types, tuples and enums is
open-coded by the Rust compiler.
10.0.6
Linkage
The Rust compiler supports various methods to link crates together both statically
and dynamically. This section will explore the various methods to link Rust
crates together, and more information about native libraries can be found in the
ffi tutorial.
In one session of compilation, the compiler can generate multiple artifacts through
the usage of either command line flags or the crate_type attribute. If one or
more command line flag is specified, all crate_type attributes will be ignored
in favor of only building the artifacts specified by command line.
--crate-type=bin, #[crate_type = "bin"] - A runnable executable will
be produced. This requires that there is a main function in the crate which
will be run when the program begins executing. This will link in all Rust
and native dependencies, producing a distributable binary.
--crate-type=lib, #[crate_type = "lib"] - A Rust library will be
produced. This is an ambiguous concept as to what exactly is produced
because a library can manifest itself in several forms. The purpose of this
generic lib option is to generate the compiler recommended style of
82
library. The output library will always be usable by rustc, but the actual
type of library may change from time-to-time. The remaining output types
are all different flavors of libraries, and the lib type can be seen as an
alias for one of them (but the actual one is compiler-defined).
--crate-type=dylib, #[crate_type = "dylib"] - A dynamic Rust library will be produced. This is different from the lib output type in
that this forces dynamic library generation. The resulting dynamic library
can be used as a dependency for other libraries and/or executables. This
output type will create *.so files on linux, *.dylib files on osx, and *.dll
files on windows.
--crate-type=staticlib, #[crate_type = "staticlib"] - A static system library will be produced. This is different from other library outputs
in that the Rust compiler will never attempt to link to staticlib outputs.
The purpose of this output type is to create a static library containing
all of the local crates code along with all upstream dependencies. The
static library is actually a *.a archive on linux and osx and a *.lib file
on windows. This format is recommended for use in situtations such as
linking Rust code into an existing non-Rust application because it will not
have dynamic dependencies on other Rust code.
--crate-type=rlib, #[crate_type = "rlib"] - A Rust library file will
be produced. This is used as an intermediate artifact and can be thought
of as a static Rust library. These rlib files, unlike staticlib files, are
interpreted by the Rust compiler in future linkage. This essentially means
that rustc will look for metadata in rlib files like it looks for metadata in
dynamic libraries. This form of output is used to produce statically linked
executables as well as staticlib outputs.
Note that these outputs are stackable in the sense that if multiple are specified,
then the compiler will produce each form of output at once without having to
recompile. However, this only applies for outputs specified by the same method.
If only crate_type attributes are specified, then they will all be built, but if one
or more --crate-type command line flag is specified, then only those outputs
will be built.
With all these different kinds of outputs, if crate A depends on crate B, then
the compiler could find B in various different forms throughout the system. The
only forms looked for by the compiler, however, are the rlib format and the
dynamic library format. With these two options for a dependent library, the
compiler must at some point make a choice between these two formats. With
this in mind, the compiler follows these rules when determining what format of
dependencies will be used:
1. If a static library is being produced, all upstream dependencies are required
to be available in rlib formats. This requirement stems from the reason
that a dynamic library cannot be converted into a static format.
83
Logging system
paths, with optional log levels. For each module containing log expressions, if
RUST_LOG contains the path to that module or a parent of that module, then
logs of the appropriate level will be output to the console.
The path to a module consists of the crate name, any parent modules, then the
module itself, all separated by double colons (::). The optional log level can
be appended to the module path with an equals sign (=) followed by the log
level, from 1 to 4, inclusive. Level 1 is the error level, 2 is warning, 3 info, and 4
debug. You can also use the symbolic constants error, warn, info, and debug.
Any logs less than or equal to the specified level will be output. If not specified
then log level 4 is assumed. Debug messages can be omitted by passing --cfg
ndebug to rustc.
As an example, to see all the logs generated by the compiler, you would set
RUST_LOG to rustc, which is the crate name (as specified in its crate_id
attribute). To narrow down the logs to just crate resolution, you would set it to
rustc::metadata::creader. To see just error logging use rustc=0.
Note that when compiling source files that dont specify a crate name the crate is
given a default name that matches the source file, with the extension removed. In
that case, to turn on logging for a program compiled from, e.g. helloworld.rs,
RUST_LOG should be set to helloworld.
Logging Expressions Rust provides several macros to log information. Heres
a simple Rust program that demonstrates all four of them:
#![feature(phase)]
#[phase(plugin, link)] extern crate log;
fn main() {
error!("This is an error log")
warn!("This is a warn log")
info!("this is an info log")
debug!("This is a debug log")
}
These four log levels correspond to levels 1-4, as controlled by RUST_LOG:
$ RUST_LOG=rust=3 ./rust
This is an error log
This is a warn log
this is an info log
11
TODO.
85
12
12.1
Influences
The essential problem that must be solved in making a fault-tolerant
software system is therefore that of fault-isolation. Different programmers will write different modules, some modules will be correct,
others will have errors. We do not want the errors in one module to
adversely affect the behaviour of a module which does not have any
errors.
Joe Armstrong
In our approach, all data is private to some process, and processes can
only communicate through communications channels. Security, as
used in this paper, is the property which guarantees that processes in
a system cannot affect each other except by explicit communication.
When security is absent, nothing which can be proven about a single
module in isolation can be guaranteed to hold when that module is
embedded in a system [. . . ]
Robert Strom and Shaula Yemini
Concurrent and applicative programming complement each other.
The ability to send messages on channels provides I/O without side
effects, while the avoidance of shared data helps keep concurrent
processes from colliding.
Rob Pike
Rust is not a particularly original language. It may however appear unusual
by contemporary standards, as its design elements are drawn from a number of
historical languages that have, with a few exceptions, fallen out of favour. Five
prominent lineages contribute the most, though their influences have come and
gone during the course of Rusts development:
The NIL (1981) and Hermes (1990) family. These languages were developed
by Robert Strom, Shaula Yemini, David Bacon and others in their group
at IBM Watson Research Center (Yorktown Heights, NY, USA).
The Erlang (1987) language, developed by Joe Armstrong, Robert Virding,
Claes Wikstrm, Mike Williams and others in their group at the Ericsson
Computer Science Laboratory (lvsj, Stockholm, Sweden) .
The Sather (1990) language, developed by Stephen Omohundro, ChuCheow Lim, Heinz Schmidt and others in their group at The International
Computer Science Institute of the University of California, Berkeley (Berkeley, CA, USA).
86
The Newsqueak (1988), Alef (1995), and Limbo (1996) family. These
languages were developed by Rob Pike, Phil Winterbottom, Sean Dorward
and others in their group at Bell Labs Computing Sciences Research Center
(Murray Hill, NJ, USA).
The Napier (1985) and Napier88 (1988) family. These languages were
developed by Malcolm Atkinson, Ron Morrison and others in their group
at the University of St. Andrews (St. Andrews, Fife, UK).
Additional specific influences can be seen from the following languages:
The structural algebraic types and compilation manager of SML.
The attribute and assembly systems of C#.
The references and deterministic destructor system of C++.
The memory region systems of the ML Kit and Cyclone.
The typeclass system of Haskell.
The lexical identifier rule of Python.
The block syntax of Ruby.
87