5 This document is the primary reference for the Rust programming language. It
6 provides three kinds of material:
8 - Chapters that informally describe each language construct and their use.
9 - Chapters that informally describe the memory model, concurrency model,
10 runtime services, linkage model and debugging facilities.
11 - Appendix chapters providing rationale and references to languages that
12 influenced the design.
14 This document does not serve as an introduction to the language. Background
15 familiarity with the language is assumed. A separate [book] is available to
16 help acquire such background familiarity.
18 This document also does not serve as a reference to the [standard] library
19 included in the language distribution. Those libraries are documented
20 separately by extracting documentation attributes from their source code. Many
21 of the features that one might expect to be language features are library
22 features in Rust, so what you're looking for may be there, not here.
24 You may also be interested in the [grammar].
26 [book]: book/index.html
27 [standard]: std/index.html
28 [grammar]: grammar.html
32 ## Unicode productions
34 A few productions in Rust's grammar permit Unicode code points outside the
35 ASCII range. We define these productions in terms of character properties
36 specified in the Unicode standard, rather than in terms of ASCII-range code
37 points. The grammar has a [Special Unicode Productions][unicodeproductions]
38 section that lists these productions.
40 [unicodeproductions]: grammar.html#special-unicode-productions
42 ## String table productions
44 Some rules in the grammar — notably [unary
45 operators](#unary-operator-expressions), [binary
46 operators](#binary-operator-expressions), and [keywords][keywords] — are
47 given in a simplified form: as a listing of a table of unquoted, printable
48 whitespace-separated strings. These cases form a subset of the rules regarding
49 the [token](#tokens) rule, and are assumed to be the result of a
50 lexical-analysis phase feeding the parser, driven by a DFA, operating over the
51 disjunction of all such string table entries.
53 [keywords]: grammar.html#keywords
55 When such a string enclosed in double-quotes (`"`) occurs inside the grammar,
56 it is an implicit reference to a single member of such a string table
57 production. See [tokens](#tokens) for more information.
63 Rust input is interpreted as a sequence of Unicode code points encoded in UTF-8.
64 Most Rust grammar rules are defined in terms of printable ASCII-range
65 code points, but a small number are defined in terms of Unicode properties or
66 explicit code point lists. [^inputformat]
68 [^inputformat]: Substitute definitions for the special Unicode productions are
69 provided to the grammar verifier, restricted to ASCII range, when verifying the
70 grammar in this document.
74 An identifier is any nonempty Unicode[^non_ascii_idents] string of the following form:
76 [^non_ascii_idents]: Non-ASCII characters in identifiers are currently feature
77 gated. This is expected to improve soon.
81 * The first character has property `XID_start`
82 * The remaining characters have property `XID_continue`
86 * The first character is `_`
87 * The identifier is more than one character, `_` alone is not an identifier
88 * The remaining characters have property `XID_continue`
90 that does _not_ occur in the set of [keywords][keywords].
92 > **Note**: `XID_start` and `XID_continue` as character properties cover the
93 > character ranges used to form the more familiar C and Java language-family
98 Comments in Rust code follow the general C++ style of line (`//`) and
99 block (`/* ... */`) comment forms. Nested block comments are supported.
101 Line comments beginning with exactly _three_ slashes (`///`), and block
102 comments beginning with exactly one repeated asterisk in the block-open
103 sequence (`/**`), are interpreted as a special syntax for `doc`
104 [attributes](#attributes). That is, they are equivalent to writing
105 `#[doc="..."]` around the body of the comment, i.e., `/// Foo` turns into
108 Line comments beginning with `//!` and block comments beginning with `/*!` are
109 doc comments that apply to the parent of the comment, rather than the item
110 that follows. That is, they are equivalent to writing `#![doc="..."]` around
111 the body of the comment. `//!` comments are usually used to document
112 modules that occupy a source file.
114 Non-doc comments are interpreted as a form of whitespace.
118 Whitespace is any non-empty string containing only the following characters:
120 - `U+0020` (space, `' '`)
121 - `U+0009` (tab, `'\t'`)
122 - `U+000A` (LF, `'\n'`)
123 - `U+000D` (CR, `'\r'`)
125 Rust is a "free-form" language, meaning that all forms of whitespace serve only
126 to separate _tokens_ in the grammar, and have no semantic significance.
128 A Rust program has identical meaning if each whitespace element is replaced
129 with any other legal whitespace element, such as a single space character.
133 Tokens are primitive productions in the grammar defined by regular
134 (non-recursive) languages. "Simple" tokens are given in [string table
135 production](#string-table-productions) form, and occur in the rest of the
136 grammar as double-quoted strings. Other tokens have exact rules given.
140 A literal is an expression consisting of a single token, rather than a sequence
141 of tokens, that immediately and directly denotes the value it evaluates to,
142 rather than referring to it by name or some other evaluation rule. A literal is
143 a form of constant expression, so is evaluated (primarily) at compile time.
147 ##### Characters and strings
149 | | Example | `#` sets | Characters | Escapes |
150 |----------------------------------------------|-----------------|------------|-------------|---------------------|
151 | [Character](#character-literals) | `'H'` | `N/A` | All Unicode | `\'` & [Byte](#byte-escapes) & [Unicode](#unicode-escapes) |
152 | [String](#string-literals) | `"hello"` | `N/A` | All Unicode | `\"` & [Byte](#byte-escapes) & [Unicode](#unicode-escapes) |
153 | [Raw](#raw-string-literals) | `r#"hello"#` | `0...` | All Unicode | `N/A` |
154 | [Byte](#byte-literals) | `b'H'` | `N/A` | All ASCII | `\'` & [Byte](#byte-escapes) |
155 | [Byte string](#byte-string-literals) | `b"hello"` | `N/A` | All ASCII | `\"` & [Byte](#byte-escapes) |
156 | [Raw byte string](#raw-byte-string-literals) | `br#"hello"#` | `0...` | All ASCII | `N/A` |
162 | `\x7F` | 8-bit character code (exactly 2 digits) |
164 | `\r` | Carriage return |
168 ##### Unicode escapes
171 | `\u{7FFF}` | 24-bit Unicode character code (up to 6 digits) |
175 | [Number literals](#number-literals)`*` | Example | Exponentiation | Suffixes |
176 |----------------------------------------|---------|----------------|----------|
177 | Decimal integer | `98_222` | `N/A` | Integer suffixes |
178 | Hex integer | `0xff` | `N/A` | Integer suffixes |
179 | Octal integer | `0o77` | `N/A` | Integer suffixes |
180 | Binary integer | `0b1111_0000` | `N/A` | Integer suffixes |
181 | Floating-point | `123.0E+77` | `Optional` | Floating-point suffixes |
183 `*` All number literals allow `_` as a visual separator: `1_234.0E+18f64`
186 | Integer | Floating-point |
187 |---------|----------------|
188 | `u8`, `i8`, `u16`, `i16`, `u32`, `i32`, `u64`, `i64`, `isize`, `usize` | `f32`, `f64` |
190 #### Character and string literals
192 ##### Character literals
194 A _character literal_ is a single Unicode character enclosed within two
195 `U+0027` (single-quote) characters, with the exception of `U+0027` itself,
196 which must be _escaped_ by a preceding `U+005C` character (`\`).
198 ##### String literals
200 A _string literal_ is a sequence of any Unicode characters enclosed within two
201 `U+0022` (double-quote) characters, with the exception of `U+0022` itself,
202 which must be _escaped_ by a preceding `U+005C` character (`\`).
204 Line-break characters are allowed in string literals. Normally they represent
205 themselves (i.e. no translation), but as a special exception, when a `U+005C`
206 character (`\`) occurs immediately before the newline, the `U+005C` character,
207 the newline, and all whitespace at the beginning of the next line are ignored.
208 Thus `a` and `b` are equal:
218 ##### Character escapes
220 Some additional _escapes_ are available in either character or non-raw string
221 literals. An escape starts with a `U+005C` (`\`) and continues with one of the
224 * An _8-bit code point escape_ starts with `U+0078` (`x`) and is
225 followed by exactly two _hex digits_. It denotes the Unicode code point
226 equal to the provided hex value.
227 * A _24-bit code point escape_ starts with `U+0075` (`u`) and is followed
228 by up to six _hex digits_ surrounded by braces `U+007B` (`{`) and `U+007D`
229 (`}`). It denotes the Unicode code point equal to the provided hex value.
230 * A _whitespace escape_ is one of the characters `U+006E` (`n`), `U+0072`
231 (`r`), or `U+0074` (`t`), denoting the Unicode values `U+000A` (LF),
232 `U+000D` (CR) or `U+0009` (HT) respectively.
233 * The _backslash escape_ is the character `U+005C` (`\`) which must be
234 escaped in order to denote *itself*.
236 ##### Raw string literals
238 Raw string literals do not process any escapes. They start with the character
239 `U+0072` (`r`), followed by zero or more of the character `U+0023` (`#`) and a
240 `U+0022` (double-quote) character. The _raw string body_ can contain any sequence
241 of Unicode characters and is terminated only by another `U+0022` (double-quote)
242 character, followed by the same number of `U+0023` (`#`) characters that preceded
243 the opening `U+0022` (double-quote) character.
245 All Unicode characters contained in the raw string body represent themselves,
246 the characters `U+0022` (double-quote) (except when followed by at least as
247 many `U+0023` (`#`) characters as were used to start the raw string literal) or
248 `U+005C` (`\`) do not have any special meaning.
250 Examples for string literals:
253 "foo"; r"foo"; // foo
254 "\"foo\""; r#""foo""#; // "foo"
257 r##"foo #"# bar"##; // foo #"# bar
259 "\x52"; "R"; r"R"; // R
260 "\\x52"; r"\x52"; // \x52
263 #### Byte and byte string literals
267 A _byte literal_ is a single ASCII character (in the `U+0000` to `U+007F`
268 range) or a single _escape_ preceded by the characters `U+0062` (`b`) and
269 `U+0027` (single-quote), and followed by the character `U+0027`. If the character
270 `U+0027` is present within the literal, it must be _escaped_ by a preceding
271 `U+005C` (`\`) character. It is equivalent to a `u8` unsigned 8-bit integer
274 ##### Byte string literals
276 A non-raw _byte string literal_ is a sequence of ASCII characters and _escapes_,
277 preceded by the characters `U+0062` (`b`) and `U+0022` (double-quote), and
278 followed by the character `U+0022`. If the character `U+0022` is present within
279 the literal, it must be _escaped_ by a preceding `U+005C` (`\`) character.
280 Alternatively, a byte string literal can be a _raw byte string literal_, defined
281 below. A byte string literal of length `n` is equivalent to a `&'static [u8; n]` borrowed fixed-sized array
282 of unsigned 8-bit integers.
284 Some additional _escapes_ are available in either byte or non-raw byte string
285 literals. An escape starts with a `U+005C` (`\`) and continues with one of the
288 * A _byte escape_ escape starts with `U+0078` (`x`) and is
289 followed by exactly two _hex digits_. It denotes the byte
290 equal to the provided hex value.
291 * A _whitespace escape_ is one of the characters `U+006E` (`n`), `U+0072`
292 (`r`), or `U+0074` (`t`), denoting the bytes values `0x0A` (ASCII LF),
293 `0x0D` (ASCII CR) or `0x09` (ASCII HT) respectively.
294 * The _backslash escape_ is the character `U+005C` (`\`) which must be
295 escaped in order to denote its ASCII encoding `0x5C`.
297 ##### Raw byte string literals
299 Raw byte string literals do not process any escapes. They start with the
300 character `U+0062` (`b`), followed by `U+0072` (`r`), followed by zero or more
301 of the character `U+0023` (`#`), and a `U+0022` (double-quote) character. The
302 _raw string body_ can contain any sequence of ASCII characters and is terminated
303 only by another `U+0022` (double-quote) character, followed by the same number of
304 `U+0023` (`#`) characters that preceded the opening `U+0022` (double-quote)
305 character. A raw byte string literal can not contain any non-ASCII byte.
307 All characters contained in the raw string body represent their ASCII encoding,
308 the characters `U+0022` (double-quote) (except when followed by at least as
309 many `U+0023` (`#`) characters as were used to start the raw string literal) or
310 `U+005C` (`\`) do not have any special meaning.
312 Examples for byte string literals:
315 b"foo"; br"foo"; // foo
316 b"\"foo\""; br#""foo""#; // "foo"
319 br##"foo #"# bar"##; // foo #"# bar
321 b"\x52"; b"R"; br"R"; // R
322 b"\\x52"; br"\x52"; // \x52
327 A _number literal_ is either an _integer literal_ or a _floating-point
328 literal_. The grammar for recognizing the two kinds of literals is mixed.
330 ##### Integer literals
332 An _integer literal_ has one of four forms:
334 * A _decimal literal_ starts with a *decimal digit* and continues with any
335 mixture of *decimal digits* and _underscores_.
336 * A _hex literal_ starts with the character sequence `U+0030` `U+0078`
337 (`0x`) and continues as any mixture of hex digits and underscores.
338 * An _octal literal_ starts with the character sequence `U+0030` `U+006F`
339 (`0o`) and continues as any mixture of octal digits and underscores.
340 * A _binary literal_ starts with the character sequence `U+0030` `U+0062`
341 (`0b`) and continues as any mixture of binary digits and underscores.
343 Like any literal, an integer literal may be followed (immediately,
344 without any spaces) by an _integer suffix_, which forcibly sets the
345 type of the literal. The integer suffix must be the name of one of the
346 integral types: `u8`, `i8`, `u16`, `i16`, `u32`, `i32`, `u64`, `i64`,
349 The type of an _unsuffixed_ integer literal is determined by type inference:
351 * If an integer type can be _uniquely_ determined from the surrounding
352 program context, the unsuffixed integer literal has that type.
354 * If the program context under-constrains the type, it defaults to the
355 signed 32-bit integer `i32`.
357 * If the program context over-constrains the type, it is considered a
360 Examples of integer literals of various forms:
367 0o70_i16; // type i16
368 0b1111_1111_1001_0000_i32; // type i32
369 0usize; // type usize
372 ##### Floating-point literals
374 A _floating-point literal_ has one of two forms:
376 * A _decimal literal_ followed by a period character `U+002E` (`.`). This is
377 optionally followed by another decimal literal, with an optional _exponent_.
378 * A single _decimal literal_ followed by an _exponent_.
380 Like integer literals, a floating-point literal may be followed by a
381 suffix, so long as the pre-suffix part does not end with `U+002E` (`.`).
382 The suffix forcibly sets the type of the literal. There are two valid
383 _floating-point suffixes_, `f32` and `f64` (the 32-bit and 64-bit floating point
384 types), which explicitly determine the type of the literal.
386 The type of an _unsuffixed_ floating-point literal is determined by
389 * If a floating-point type can be _uniquely_ determined from the
390 surrounding program context, the unsuffixed floating-point literal
393 * If the program context under-constrains the type, it defaults to `f64`.
395 * If the program context over-constrains the type, it is considered a
398 Examples of floating-point literals of various forms:
401 123.0f64; // type f64
404 12E+99_f64; // type f64
405 let x: f64 = 2.; // type f64
408 This last example is different because it is not possible to use the suffix
409 syntax with a floating point literal ending in a period. `2.f64` would attempt
410 to call a method named `f64` on `2`.
412 The representation semantics of floating-point numbers are described in
413 ["Machine Types"](#machine-types).
415 #### Boolean literals
417 The two values of the boolean type are written `true` and `false`.
421 Symbols are a general class of printable [tokens](#tokens) that play structural
422 roles in a variety of grammar productions. They are a
423 set of remaining miscellaneous printable tokens that do not
424 otherwise appear as [unary operators](#unary-operator-expressions), [binary
425 operators](#binary-operator-expressions), or [keywords][keywords].
426 They are catalogued in [the Symbols section][symbols] of the Grammar document.
428 [symbols]: grammar.html#symbols
433 A _path_ is a sequence of one or more path components _logically_ separated by
434 a namespace qualifier (`::`). If a path consists of only one component, it may
435 refer to either an [item](#items) or a [variable](#variables) in a local control
436 scope. If a path has multiple components, it refers to an item.
438 Every item has a _canonical path_ within its crate, but the path naming an item
439 is only meaningful within a given crate. There is no global namespace across
440 crates; an item's canonical path merely identifies it within the crate.
442 Two examples of simple paths consisting of only identifier components:
449 Path components are usually [identifiers](#identifiers), but they may
450 also include angle-bracket-enclosed lists of type arguments. In
451 [expression](#expressions) context, the type argument list is given
452 after a `::` namespace qualifier in order to disambiguate it from a
453 relational expression involving the less-than symbol (`<`). In type
454 expression context, the final namespace qualifier is omitted.
456 Two examples of paths with type arguments:
459 # struct HashMap<K, V>(K,V);
461 # fn id<T>(t: T) -> T { t }
462 type T = HashMap<i32,String>; // Type arguments used in a type expression
463 let x = id::<i32>(10); // Type arguments used in a call expression
467 Paths can be denoted with various leading qualifiers to change the meaning of
470 * Paths starting with `::` are considered to be global paths where the
471 components of the path start being resolved from the crate root. Each
472 identifier in the path must resolve to an item.
480 ::a::foo(); // call a's foo function
486 * Paths starting with the keyword `super` begin resolution relative to the
487 parent module. Each further identifier must resolve to an item.
495 super::a::foo(); // call a's foo function
501 * Paths starting with the keyword `self` begin resolution relative to the
502 current module. Each further identifier must resolve to an item.
514 A number of minor features of Rust are not central enough to have their own
515 syntax, and yet are not implementable as functions. Instead, they are given
516 names, and invoked through a consistent syntax: `some_extension!(...)`.
518 Users of `rustc` can define new syntax extensions in two ways:
520 * [Compiler plugins][plugin] can include arbitrary Rust code that
521 manipulates syntax trees at compile time. Note that the interface
522 for compiler plugins is considered highly unstable.
524 * [Macros](book/macros.html) define new syntax in a higher-level,
529 `macro_rules` allows users to define syntax extension in a declarative way. We
530 call such extensions "macros by example" or simply "macros" — to be distinguished
531 from the "procedural macros" defined in [compiler plugins][plugin].
533 Currently, macros can expand to expressions, statements, items, or patterns.
535 (A `sep_token` is any token other than `*` and `+`. A `non_special_token` is
536 any token other than a delimiter or `$`.)
538 The macro expander looks up macro invocations by name, and tries each macro
539 rule in turn. It transcribes the first successful match. Matching and
540 transcription are closely related to each other, and we will describe them
545 The macro expander matches and transcribes every token that does not begin with
546 a `$` literally, including delimiters. For parsing reasons, delimiters must be
547 balanced, but they are otherwise not special.
549 In the matcher, `$` _name_ `:` _designator_ matches the nonterminal in the Rust
550 syntax named by _designator_. Valid designators are:
552 * `item`: an [item](#items)
553 * `block`: a [block](#block-expressions)
554 * `stmt`: a [statement](#statements)
555 * `pat`: a [pattern](#match-expressions)
556 * `expr`: an [expression](#expressions)
557 * `ty`: a [type](#types)
558 * `ident`: an [identifier](#identifiers)
559 * `path`: a [path](#paths)
560 * `tt`: either side of the `=>` in macro rules
561 * `meta`: the contents of an [attribute](#attributes)
563 In the transcriber, the
564 designator is already known, and so only the name of a matched nonterminal comes
565 after the dollar sign.
567 In both the matcher and transcriber, the Kleene star-like operator indicates
568 repetition. The Kleene star operator consists of `$` and parentheses, optionally
569 followed by a separator token, followed by `*` or `+`. `*` means zero or more
570 repetitions, `+` means at least one repetition. The parentheses are not matched or
571 transcribed. On the matcher side, a name is bound to _all_ of the names it
572 matches, in a structure that mimics the structure of the repetition encountered
573 on a successful match. The job of the transcriber is to sort that structure
576 The rules for transcription of these repetitions are called "Macro By Example".
577 Essentially, one "layer" of repetition is discharged at a time, and all of them
578 must be discharged by the time a name is transcribed. Therefore, `( $( $i:ident
579 ),* ) => ( $i )` is an invalid macro, but `( $( $i:ident ),* ) => ( $( $i:ident
580 ),* )` is acceptable (if trivial).
582 When Macro By Example encounters a repetition, it examines all of the `$`
583 _name_ s that occur in its body. At the "current layer", they all must repeat
584 the same number of times, so ` ( $( $i:ident ),* ; $( $j:ident ),* ) => ( $(
585 ($i,$j) ),* )` is valid if given the argument `(a,b,c ; d,e,f)`, but not
586 `(a,b,c ; d,e)`. The repetition walks through the choices at that layer in
587 lockstep, so the former input transcribes to `(a,d), (b,e), (c,f)`.
589 Nested repetitions are allowed.
591 ### Parsing limitations
593 The parser used by the macro system is reasonably powerful, but the parsing of
594 Rust syntax is restricted in two ways:
596 1. Macro definitions are required to include suitable separators after parsing
597 expressions and other bits of the Rust grammar. This implies that
598 a macro definition like `$i:expr [ , ]` is not legal, because `[` could be part
599 of an expression. A macro definition like `$i:expr,` or `$i:expr;` would be legal,
600 however, because `,` and `;` are legal separators. See [RFC 550] for more information.
601 2. The parser must have eliminated all ambiguity by the time it reaches a `$`
602 _name_ `:` _designator_. This requirement most often affects name-designator
603 pairs when they occur at the beginning of, or immediately after, a `$(...)*`;
604 requiring a distinctive token in front can solve the problem.
606 [RFC 550]: https://github.com/rust-lang/rfcs/blob/master/text/0550-macro-future-proofing.md
608 # Crates and source files
610 Although Rust, like any other language, can be implemented by an interpreter as
611 well as a compiler, the only existing implementation is a compiler,
613 always been designed to be compiled. For these reasons, this section assumes a
616 Rust's semantics obey a *phase distinction* between compile-time and
617 run-time.[^phase-distinction] Semantic rules that have a *static
618 interpretation* govern the success or failure of compilation, while
620 that have a *dynamic interpretation* govern the behavior of the program at
623 [^phase-distinction]: This distinction would also exist in an interpreter.
624 Static checks like syntactic analysis, type checking, and lints should
625 happen before the program is executed regardless of when it is executed.
627 The compilation model centers on artifacts called _crates_. Each compilation
628 processes a single crate in source form, and if successful, produces a single
629 crate in binary form: either an executable or some sort of
630 library.[^cratesourcefile]
632 [^cratesourcefile]: A crate is somewhat analogous to an *assembly* in the
633 ECMA-335 CLI model, a *library* in the SML/NJ Compilation Manager, a *unit*
634 in the Owens and Flatt module system, or a *configuration* in Mesa.
636 A _crate_ is a unit of compilation and linking, as well as versioning,
637 distribution and runtime loading. A crate contains a _tree_ of nested
638 [module](#modules) scopes. The top level of this tree is a module that is
639 anonymous (from the point of view of paths within the module) and any item
640 within a crate has a canonical [module path](#paths) denoting its location
641 within the crate's module tree.
643 The Rust compiler is always invoked with a single source file as input, and
644 always produces a single output crate. The processing of that source file may
645 result in other source files being loaded as modules. Source files have the
648 A Rust source file describes a module, the name and location of which —
649 in the module tree of the current crate — are defined from outside the
650 source file: either by an explicit `mod_item` in a referencing source file, or
651 by the name of the crate itself. Every source file is a module, but not every
652 module needs its own source file: [module definitions](#modules) can be nested
655 Each source file contains a sequence of zero or more `item` definitions, and
656 may optionally begin with any number of [attributes](#items-and-attributes)
657 that apply to the containing module, most of which influence the behavior of
658 the compiler. The anonymous crate module can have additional attributes that
659 apply to the crate as a whole.
662 // Specify the crate name.
663 #![crate_name = "projx"]
665 // Specify the type of output artifact.
666 #![crate_type = "lib"]
668 // Turn on a warning.
669 // This can be done in any module, not just the anonymous crate module.
670 #![warn(non_camel_case_types)]
673 A crate that contains a `main` function can be compiled to an executable. If a
674 `main` function is present, its return type must be [`unit`](#tuple-types)
675 and it must take no arguments.
677 # Items and attributes
679 Crates contain [items](#items), each of which may have some number of
680 [attributes](#attributes) attached to it.
684 An _item_ is a component of a crate. Items are organized within a crate by a
685 nested set of [modules](#modules). Every crate has a single "outermost"
686 anonymous module; all further items within the crate have [paths](#paths)
687 within the module tree of the crate.
689 Items are entirely determined at compile-time, generally remain fixed during
690 execution, and may reside in read-only memory.
692 There are several kinds of item:
694 * [`extern crate` declarations](#extern-crate-declarations)
695 * [`use` declarations](#use-declarations)
696 * [modules](#modules)
697 * [functions](#functions)
698 * [type definitions](grammar.html#type-definitions)
699 * [structs](#structs)
700 * [enumerations](#enumerations)
701 * [constant items](#constant-items)
702 * [static items](#static-items)
704 * [implementations](#implementations)
706 Some items form an implicit scope for the declaration of sub-items. In other
707 words, within a function or module, declarations of items can (in many cases)
708 be mixed with the statements, control blocks, and similar artifacts that
709 otherwise compose the item body. The meaning of these scoped items is the same
710 as if the item was declared outside the scope — it is still a static item
711 — except that the item's *path name* within the module namespace is
712 qualified by the name of the enclosing item, or is private to the enclosing
713 item (in the case of functions). The grammar specifies the exact locations in
714 which sub-item declarations may appear.
718 All items except modules, constants and statics may be *parameterized* by type.
719 Type parameters are given as a comma-separated list of identifiers enclosed in
720 angle brackets (`<...>`), after the name of the item and before its definition.
721 The type parameters of an item are considered "part of the name", not part of
722 the type of the item. A referencing [path](#paths) must (in principle) provide
723 type arguments as a list of comma-separated types enclosed within angle
724 brackets, in order to refer to the type-parameterized item. In practice, the
725 type-inference system can usually infer such argument types from context. There
726 are no general type-parametric types, only type-parametric items. That is, Rust
727 has no notion of type abstraction: there are no higher-ranked (or "forall") types
728 abstracted over other types, though higher-ranked types do exist for lifetimes.
732 A module is a container for zero or more [items](#items).
734 A _module item_ is a module, surrounded in braces, named, and prefixed with the
735 keyword `mod`. A module item introduces a new, named module into the tree of
736 modules making up a crate. Modules can nest arbitrarily.
738 An example of a module:
742 type Complex = (f64, f64);
743 fn sin(f: f64) -> f64 {
747 fn cos(f: f64) -> f64 {
751 fn tan(f: f64) -> f64 {
758 Modules and types share the same namespace. Declaring a named type with
759 the same name as a module in scope is forbidden: that is, a type definition,
760 trait, struct, enumeration, or type parameter can't shadow the name of a module
761 in scope, or vice versa.
763 A module without a body is loaded from an external file, by default with the
764 same name as the module, plus the `.rs` extension. When a nested submodule is
765 loaded from an external file, it is loaded from a subdirectory path that
766 mirrors the module hierarchy.
769 // Load the `vec` module from `vec.rs`
773 // Load the `local_data` module from `thread/local_data.rs`
774 // or `thread/local_data/mod.rs`.
779 The directories and files used for loading external file modules can be
780 influenced with the `path` attribute.
783 #[path = "thread_files"]
785 // Load the `local_data` module from `thread_files/tls.rs`
791 #### Extern crate declarations
793 An _`extern crate` declaration_ specifies a dependency on an external crate.
794 The external crate is then bound into the declaring scope as the `ident`
795 provided in the `extern_crate_decl`.
797 The external crate is resolved to a specific `soname` at compile time, and a
798 runtime linkage requirement to that `soname` is passed to the linker for
799 loading at runtime. The `soname` is resolved at compile time by scanning the
800 compiler's library path and matching the optional `crateid` provided against
801 the `crateid` attributes that were declared on the external crate when it was
802 compiled. If no `crateid` is provided, a default `name` attribute is assumed,
803 equal to the `ident` given in the `extern_crate_decl`.
805 Three examples of `extern crate` declarations:
810 extern crate std; // equivalent to: extern crate std as std;
812 extern crate std as ruststd; // linking to 'std' under another name
815 #### Use declarations
817 A _use declaration_ creates one or more local name bindings synonymous with
818 some other [path](#paths). Usually a `use` declaration is used to shorten the
819 path required to refer to a module item. These declarations may appear at the
820 top of [modules](#modules) and [blocks](grammar.html#block-expressions).
822 > **Note**: Unlike in many languages,
823 > `use` declarations in Rust do *not* declare linkage dependency with external crates.
824 > Rather, [`extern crate` declarations](#extern-crate-declarations) declare linkage dependencies.
826 Use declarations support a number of convenient shortcuts:
828 * Rebinding the target name as a new local name, using the syntax `use p::q::r as x;`
829 * Simultaneously binding a list of paths differing only in their final element,
830 using the glob-like brace syntax `use a::b::{c,d,e,f};`
831 * Binding all paths matching a given prefix, using the asterisk wildcard syntax
833 * Simultaneously binding a list of paths differing only in their final element
834 and their immediate parent module, using the `self` keyword, such as
835 `use a::b::{self, c, d};`
837 An example of `use` declarations:
840 use std::option::Option::{Some, None};
841 use std::collections::hash_map::{self, HashMap};
844 fn bar(map1: HashMap<String, usize>, map2: hash_map::HashMap<String, usize>){}
847 // Equivalent to 'foo(vec![std::option::Option::Some(1.0f64),
848 // std::option::Option::None]);'
849 foo(vec![Some(1.0f64), None]);
851 // Both `hash_map` and `HashMap` are in scope.
852 let map1 = HashMap::new();
853 let map2 = hash_map::HashMap::new();
858 Like items, `use` declarations are private to the containing module, by
859 default. Also like items, a `use` declaration can be public, if qualified by
860 the `pub` keyword. Such a `use` declaration serves to _re-export_ a name. A
861 public `use` declaration can therefore _redirect_ some public name to a
862 different target definition: even a definition with a private canonical path,
863 inside a different module. If a sequence of such redirections form a cycle or
864 cannot be resolved unambiguously, they represent a compile-time error.
866 An example of re-exporting:
871 pub use quux::foo::{bar, baz};
880 In this example, the module `quux` re-exports two public names defined in
883 Also note that the paths contained in `use` items are relative to the crate
884 root. So, in the previous example, the `use` refers to `quux::foo::{bar,
885 baz}`, and not simply to `foo::{bar, baz}`. This also means that top-level
886 module declarations should be at the crate root if direct usage of the declared
887 modules within `use` items is desired. It is also possible to use `self` and
888 `super` at the beginning of a `use` item to refer to the current and direct
889 parent modules respectively. All rules regarding accessing declared modules in
890 `use` declarations apply to both module declarations and `extern crate`
893 An example of what will and will not work for `use` items:
896 # #![allow(unused_imports)]
897 use foo::baz::foobaz; // good: foo is at the root of the crate
905 use foo::example::iter; // good: foo is at crate root
906 // use example::iter; // bad: example is not at the crate root
907 use self::baz::foobaz; // good: self refers to module 'foo'
908 use foo::bar::foobar; // good: foo is at crate root
915 use super::bar::foobar; // good: super refers to module 'foo'
925 A _function item_ defines a sequence of [statements](#statements) and a
926 final [expression](#expressions), along with a name and a set of
927 parameters. Other than a name, all these are optional.
928 Functions are declared with the keyword `fn`. Functions may declare a
929 set of *input* [*variables*](#variables) as parameters, through which the caller
930 passes arguments into the function, and the *output* [*type*](#types)
931 of the value the function will return to its caller on completion.
933 A function may also be copied into a first-class *value*, in which case the
934 value has the corresponding [*function type*](#function-types), and can be used
935 otherwise exactly as a function item (with a minor additional cost of calling
936 the function indirectly).
938 Every control path in a function logically ends with a `return` expression or a
939 diverging expression. If the outermost block of a function has a
940 value-producing expression in its final-expression position, that expression is
941 interpreted as an implicit `return` expression applied to the final-expression.
943 An example of a function:
946 fn add(x: i32, y: i32) -> i32 {
951 As with `let` bindings, function arguments are irrefutable patterns, so any
952 pattern that is valid in a let binding is also valid as an argument.
955 fn first((value, _): (i32, i32)) -> i32 { value }
959 #### Generic functions
961 A _generic function_ allows one or more _parameterized types_ to appear in its
962 signature. Each type parameter must be explicitly declared, in an
963 angle-bracket-enclosed, comma-separated list following the function name.
966 // foo is generic over A and B
968 fn foo<A, B>(x: A, y: B) {
971 Inside the function signature and body, the name of the type parameter can be
972 used as a type name. [Trait](#traits) bounds can be specified for type parameters
973 to allow methods with that trait to be called on values of that type. This is
974 specified using the `where` syntax:
977 fn foo<T>(x: T) where T: Debug {
980 When a generic function is referenced, its type is instantiated based on the
981 context of the reference. For example, calling the `foo` function here:
986 fn foo<T>(x: &[T]) where T: Debug {
994 will instantiate type parameter `T` with `i32`.
996 The type parameters can also be explicitly supplied in a trailing
997 [path](#paths) component after the function name. This might be necessary if
998 there is not sufficient context to determine the type parameters. For example,
999 `mem::size_of::<u32>() == 4`.
1001 #### Diverging functions
1003 A special kind of function can be declared with a `!` character where the
1004 output type would normally be. For example:
1007 fn my_err(s: &str) -> ! {
1013 We call such functions "diverging" because they never return a value to the
1014 caller. Every control path in a diverging function must end with a `panic!()` or
1015 a call to another diverging function on every control path. The `!` annotation
1016 does *not* denote a type.
1018 It might be necessary to declare a diverging function because as mentioned
1019 previously, the typechecker checks that every control path in a function ends
1020 with a [`return`](#return-expressions) or diverging expression. So, if `my_err`
1021 were declared without the `!` annotation, the following code would not
1025 # fn my_err(s: &str) -> ! { panic!() }
1027 fn f(i: i32) -> i32 {
1032 my_err("Bad number!");
1037 This will not compile without the `!` annotation on `my_err`, since the `else`
1038 branch of the conditional in `f` does not return an `i32`, as required by the
1039 signature of `f`. Adding the `!` annotation to `my_err` informs the
1040 typechecker that, should control ever enter `my_err`, no further type judgments
1041 about `f` need to hold, since control will never resume in any context that
1042 relies on those judgments. Thus the return type on `f` only needs to reflect
1043 the `if` branch of the conditional.
1045 #### Extern functions
1047 Extern functions are part of Rust's foreign function interface, providing the
1048 opposite functionality to [external blocks](#external-blocks). Whereas
1049 external blocks allow Rust code to call foreign code, extern functions with
1050 bodies defined in Rust code _can be called by foreign code_. They are defined
1051 in the same way as any other Rust function, except that they have the `extern`
1055 // Declares an extern fn, the ABI defaults to "C"
1056 extern fn new_i32() -> i32 { 0 }
1058 // Declares an extern fn with "stdcall" ABI
1059 extern "stdcall" fn new_i32_stdcall() -> i32 { 0 }
1062 Unlike normal functions, extern fns have type `extern "ABI" fn()`. This is the
1063 same type as the functions declared in an extern block.
1066 # extern fn new_i32() -> i32 { 0 }
1067 let fptr: extern "C" fn() -> i32 = new_i32;
1070 Extern functions may be called directly from Rust code as Rust uses large,
1071 contiguous stack segments like C.
1075 A _type alias_ defines a new name for an existing [type](#types). Type
1076 aliases are declared with the keyword `type`. Every value has a single,
1077 specific type, but may implement several different traits, or be compatible with
1078 several different type constraints.
1080 For example, the following defines the type `Point` as a synonym for the type
1081 `(u8, u8)`, the type of pairs of unsigned 8 bit integers:
1084 type Point = (u8, u8);
1085 let p: Point = (41, 68);
1090 A _struct_ is a nominal [struct type](#struct-types) defined with the
1093 An example of a `struct` item and its use:
1096 struct Point {x: i32, y: i32}
1097 let p = Point {x: 10, y: 11};
1101 A _tuple struct_ is a nominal [tuple type](#tuple-types), also defined with
1102 the keyword `struct`. For example:
1105 struct Point(i32, i32);
1106 let p = Point(10, 11);
1107 let px: i32 = match p { Point(x, _) => x };
1110 A _unit-like struct_ is a struct without any fields, defined by leaving off
1111 the list of fields entirely. Such a struct implicitly defines a constant of
1112 its type with the same name. For example:
1115 # #![feature(braced_empty_structs)]
1117 let c = [Cookie, Cookie {}, Cookie, Cookie {}];
1123 # #![feature(braced_empty_structs)]
1125 const Cookie: Cookie = Cookie {};
1126 let c = [Cookie, Cookie {}, Cookie, Cookie {}];
1129 The precise memory layout of a struct is not specified. One can specify a
1130 particular layout using the [`repr` attribute](#ffi-attributes).
1134 An _enumeration_ is a simultaneous definition of a nominal [enumerated
1135 type](#enumerated-types) as well as a set of *constructors*, that can be used
1136 to create or pattern-match values of the corresponding enumerated type.
1138 Enumerations are declared with the keyword `enum`.
1140 An example of an `enum` item and its use:
1148 let mut a: Animal = Animal::Dog;
1152 Enumeration constructors can have either named or unnamed fields:
1157 Cat { name: String, weight: f64 }
1160 let mut a: Animal = Animal::Dog("Cocoa".to_string(), 37.2);
1161 a = Animal::Cat { name: "Spotty".to_string(), weight: 2.7 };
1164 In this example, `Cat` is a _struct-like enum variant_,
1165 whereas `Dog` is simply called an enum variant.
1167 Enums have a discriminant. You can assign them explicitly:
1175 If a discriminant isn't assigned, they start at zero, and add one for each
1178 You can cast an enum to get this value:
1181 # enum Foo { Bar = 123 }
1182 let x = Foo::Bar as u32; // x is now 123u32
1185 This only works as long as none of the variants have data attached. If
1186 it were `Bar(i32)`, this is disallowed.
1190 A *constant item* is a named _constant value_ which is not associated with a
1191 specific memory location in the program. Constants are essentially inlined
1192 wherever they are used, meaning that they are copied directly into the relevant
1193 context when used. References to the same constant are not necessarily
1194 guaranteed to refer to the same memory address.
1196 Constant values must not have destructors, and otherwise permit most forms of
1197 data. Constants may refer to the address of other constants, in which case the
1198 address will have the `static` lifetime. The compiler is, however, still at
1199 liberty to translate the constant many times, so the address referred to may not
1202 Constants must be explicitly typed. The type may be `bool`, `char`, a number, or
1203 a type derived from those primitive types. The derived types are references with
1204 the `static` lifetime, fixed-size arrays, tuples, enum variants, and structs.
1207 const BIT1: u32 = 1 << 0;
1208 const BIT2: u32 = 1 << 1;
1210 const BITS: [u32; 2] = [BIT1, BIT2];
1211 const STRING: &'static str = "bitstring";
1213 struct BitsNStrings<'a> {
1218 const BITS_N_STRINGS: BitsNStrings<'static> = BitsNStrings {
1226 A *static item* is similar to a *constant*, except that it represents a precise
1227 memory location in the program. A static is never "inlined" at the usage site,
1228 and all references to it refer to the same memory location. Static items have
1229 the `static` lifetime, which outlives all other lifetimes in a Rust program.
1230 Static items may be placed in read-only memory if they do not contain any
1231 interior mutability.
1233 Statics may contain interior mutability through the `UnsafeCell` language item.
1234 All access to a static is safe, but there are a number of restrictions on
1237 * Statics may not contain any destructors.
1238 * The types of static values must ascribe to `Sync` to allow thread-safe access.
1239 * Statics may not refer to other statics by value, only by reference.
1240 * Constants cannot refer to statics.
1242 Constants should in general be preferred over statics, unless large amounts of
1243 data are being stored, or single-address and mutability properties are required.
1245 #### Mutable statics
1247 If a static item is declared with the `mut` keyword, then it is allowed to
1248 be modified by the program. One of Rust's goals is to make concurrency bugs
1249 hard to run into, and this is obviously a very large source of race conditions
1250 or other bugs. For this reason, an `unsafe` block is required when either
1251 reading or writing a mutable static variable. Care should be taken to ensure
1252 that modifications to a mutable static are safe with respect to other threads
1253 running in the same process.
1255 Mutable statics are still very useful, however. They can be used with C
1256 libraries and can also be bound from C libraries (in an `extern` block).
1259 # fn atomic_add(_: &mut u32, _: u32) -> u32 { 2 }
1261 static mut LEVELS: u32 = 0;
1263 // This violates the idea of no shared state, and this doesn't internally
1264 // protect against races, so this function is `unsafe`
1265 unsafe fn bump_levels_unsafe1() -> u32 {
1271 // Assuming that we have an atomic_add function which returns the old value,
1272 // this function is "safe" but the meaning of the return value may not be what
1273 // callers expect, so it's still marked as `unsafe`
1274 unsafe fn bump_levels_unsafe2() -> u32 {
1275 return atomic_add(&mut LEVELS, 1);
1279 Mutable statics have the same restrictions as normal statics, except that the
1280 type of the value is not required to ascribe to `Sync`.
1284 A _trait_ describes an abstract interface that types can
1285 implement. This interface consists of associated items, which come in
1292 Associated functions whose first parameter is named `self` are called
1293 methods and may be invoked using `.` notation (e.g., `x.foo()`).
1295 All traits define an implicit type parameter `Self` that refers to
1296 "the type that is implementing this interface". Traits may also
1297 contain additional type parameters. These type parameters (including
1298 `Self`) may be constrained by other traits and so forth as usual.
1300 Trait bounds on `Self` are considered "supertraits". These are
1301 required to be acyclic. Supertraits are somewhat different from other
1302 constraints in that they affect what methods are available in the
1303 vtable when the trait is used as a [trait object](#trait-objects).
1305 Traits are implemented for specific types through separate
1306 [implementations](#implementations).
1308 Consider the following trait:
1311 # type Surface = i32;
1312 # type BoundingBox = i32;
1314 fn draw(&self, Surface);
1315 fn bounding_box(&self) -> BoundingBox;
1319 This defines a trait with two methods. All values that have
1320 [implementations](#implementations) of this trait in scope can have their
1321 `draw` and `bounding_box` methods called, using `value.bounding_box()`
1322 [syntax](#method-call-expressions).
1324 Traits can include default implementations of methods, as in:
1329 fn baz(&self) { println!("We called baz."); }
1333 Here the `baz` method has a default implementation, so types that implement
1334 `Foo` need only implement `bar`. It is also possible for implementing types
1335 to override a method that has a default implementation.
1337 Type parameters can be specified for a trait to make it generic. These appear
1338 after the trait name, using the same syntax used in [generic
1339 functions](#generic-functions).
1343 fn len(&self) -> u32;
1344 fn elt_at(&self, n: u32) -> T;
1345 fn iter<F>(&self, F) where F: Fn(T);
1349 It is also possible to define associated types for a trait. Consider the
1350 following example of a `Container` trait. Notice how the type is available
1351 for use in the method signatures:
1357 fn insert(&mut self, Self::E);
1361 In order for a type to implement this trait, it must not only provide
1362 implementations for every method, but it must specify the type `E`. Here's
1363 an implementation of `Container` for the standard library type `Vec`:
1368 # fn empty() -> Self;
1369 # fn insert(&mut self, Self::E);
1371 impl<T> Container for Vec<T> {
1373 fn empty() -> Vec<T> { Vec::new() }
1374 fn insert(&mut self, x: T) { self.push(x); }
1378 Generic functions may use traits as _bounds_ on their type parameters. This
1379 will have two effects:
1381 - Only types that have the trait may instantiate the parameter.
1382 - Within the generic function, the methods of the trait can be
1383 called on values that have the parameter's type.
1388 # type Surface = i32;
1389 # trait Shape { fn draw(&self, Surface); }
1390 fn draw_twice<T: Shape>(surface: Surface, sh: T) {
1396 Traits also define a [trait object](#trait-objects) with the same
1397 name as the trait. Values of this type are created by coercing from a
1398 pointer of some specific type to a pointer of trait type. For example,
1399 `&T` could be coerced to `&Shape` if `T: Shape` holds (and similarly
1400 for `Box<T>`). This coercion can either be implicit or
1401 [explicit](#type-cast-expressions). Here is an example of an explicit
1406 impl Shape for i32 { }
1407 let mycircle = 0i32;
1408 let myshape: Box<Shape> = Box::new(mycircle) as Box<Shape>;
1411 The resulting value is a box containing the value that was cast, along with
1412 information that identifies the methods of the implementation that was used.
1413 Values with a trait type can have [methods called](#method-call-expressions) on
1414 them, for any method in the trait, and can be used to instantiate type
1415 parameters that are bounded by the trait.
1417 Trait methods may be static, which means that they lack a `self` argument.
1418 This means that they can only be called with function call syntax (`f(x)`) and
1419 not method call syntax (`obj.f()`). The way to refer to the name of a static
1420 method is to qualify it with the trait name, treating the trait name like a
1421 module. For example:
1425 fn from_i32(n: i32) -> Self;
1428 fn from_i32(n: i32) -> f64 { n as f64 }
1430 let x: f64 = Num::from_i32(42);
1433 Traits may inherit from other traits. Consider the following example:
1436 trait Shape { fn area(&self) -> f64; }
1437 trait Circle : Shape { fn radius(&self) -> f64; }
1440 The syntax `Circle : Shape` means that types that implement `Circle` must also
1441 have an implementation for `Shape`. Multiple supertraits are separated by `+`,
1442 `trait Circle : Shape + PartialEq { }`. In an implementation of `Circle` for a
1443 given type `T`, methods can refer to `Shape` methods, since the typechecker
1444 checks that any type with an implementation of `Circle` also has an
1445 implementation of `Shape`:
1450 trait Shape { fn area(&self) -> f64; }
1451 trait Circle : Shape { fn radius(&self) -> f64; }
1452 impl Shape for Foo {
1453 fn area(&self) -> f64 {
1457 impl Circle for Foo {
1458 fn radius(&self) -> f64 {
1459 println!("calling area: {}", self.area());
1469 In type-parameterized functions, methods of the supertrait may be called on
1470 values of subtrait-bound type parameters. Referring to the previous example of
1471 `trait Circle : Shape`:
1474 # trait Shape { fn area(&self) -> f64; }
1475 # trait Circle : Shape { fn radius(&self) -> f64; }
1476 fn radius_times_area<T: Circle>(c: T) -> f64 {
1477 // `c` is both a Circle and a Shape
1478 c.radius() * c.area()
1482 Likewise, supertrait methods may also be called on trait objects.
1485 # trait Shape { fn area(&self) -> f64; }
1486 # trait Circle : Shape { fn radius(&self) -> f64; }
1487 # impl Shape for i32 { fn area(&self) -> f64 { 0.0 } }
1488 # impl Circle for i32 { fn radius(&self) -> f64 { 0.0 } }
1489 # let mycircle = 0i32;
1490 let mycircle = Box::new(mycircle) as Box<Circle>;
1491 let nonsense = mycircle.radius() * mycircle.area();
1496 An _implementation_ is an item that implements a [trait](#traits) for a
1499 Implementations are defined with the keyword `impl`.
1502 # #[derive(Copy, Clone)]
1503 # struct Point {x: f64, y: f64};
1504 # type Surface = i32;
1505 # struct BoundingBox {x: f64, y: f64, width: f64, height: f64};
1506 # trait Shape { fn draw(&self, Surface); fn bounding_box(&self) -> BoundingBox; }
1507 # fn do_draw_circle(s: Surface, c: Circle) { }
1513 impl Copy for Circle {}
1515 impl Clone for Circle {
1516 fn clone(&self) -> Circle { *self }
1519 impl Shape for Circle {
1520 fn draw(&self, s: Surface) { do_draw_circle(s, *self); }
1521 fn bounding_box(&self) -> BoundingBox {
1522 let r = self.radius;
1524 x: self.center.x - r,
1525 y: self.center.y - r,
1533 It is possible to define an implementation without referring to a trait. The
1534 methods in such an implementation can only be used as direct calls on the values
1535 of the type that the implementation targets. In such an implementation, the
1536 trait type and `for` after `impl` are omitted. Such implementations are limited
1537 to nominal types (enums, structs, trait objects), and the implementation must
1538 appear in the same crate as the `self` type:
1541 struct Point {x: i32, y: i32}
1545 println!("Point is at ({}, {})", self.x, self.y);
1549 let my_point = Point {x: 10, y:11};
1553 When a trait _is_ specified in an `impl`, all methods declared as part of the
1554 trait must be implemented, with matching types and type parameter counts.
1556 An implementation can take type parameters, which can be different from the
1557 type parameters taken by the trait it implements. Implementation parameters
1558 are written after the `impl` keyword.
1561 # trait Seq<T> { fn dummy(&self, _: T) { } }
1562 impl<T> Seq<T> for Vec<T> {
1565 impl Seq<bool> for u32 {
1566 /* Treat the integer as a sequence of bits */
1572 External blocks form the basis for Rust's foreign function interface.
1573 Declarations in an external block describe symbols in external, non-Rust
1576 Functions within external blocks are declared in the same way as other Rust
1577 functions, with the exception that they may not have a body and are instead
1578 terminated by a semicolon.
1580 Functions within external blocks may be called by Rust code, just like
1581 functions defined in Rust. The Rust compiler automatically translates between
1582 the Rust ABI and the foreign ABI.
1584 A number of [attributes](#attributes) control the behavior of external blocks.
1586 By default external blocks assume that the library they are calling uses the
1587 standard C "cdecl" ABI. Other ABIs may be specified using an `abi` string, as
1591 // Interface to the Windows API
1592 extern "stdcall" { }
1595 The `link` attribute allows the name of the library to be specified. When
1596 specified the compiler will attempt to link against the native library of the
1600 #[link(name = "crypto")]
1604 The type of a function declared in an extern block is `extern "abi" fn(A1, ...,
1605 An) -> R`, where `A1...An` are the declared types of its arguments and `R` is
1606 the declared return type.
1608 It is valid to add the `link` attribute on an empty extern block. You can use
1609 this to satisfy the linking requirements of extern blocks elsewhere in your code
1610 (including upstream crates) instead of adding the attribute to each extern block.
1612 ## Visibility and Privacy
1614 These two terms are often used interchangeably, and what they are attempting to
1615 convey is the answer to the question "Can this item be used at this location?"
1617 Rust's name resolution operates on a global hierarchy of namespaces. Each level
1618 in the hierarchy can be thought of as some item. The items are one of those
1619 mentioned above, but also include external crates. Declaring or defining a new
1620 module can be thought of as inserting a new tree into the hierarchy at the
1621 location of the definition.
1623 To control whether interfaces can be used across modules, Rust checks each use
1624 of an item to see whether it should be allowed or not. This is where privacy
1625 warnings are generated, or otherwise "you used a private item of another module
1626 and weren't allowed to."
1628 By default, everything in Rust is *private*, with one exception. Enum variants
1629 in a `pub` enum are also public by default. When an item is declared as `pub`,
1630 it can be thought of as being accessible to the outside world. For example:
1634 // Declare a private struct
1637 // Declare a public struct with a private field
1642 // Declare a public enum with two public variants
1644 PubliclyAccessibleState,
1645 PubliclyAccessibleState2,
1649 With the notion of an item being either public or private, Rust allows item
1650 accesses in two cases:
1652 1. If an item is public, then it can be used externally through any of its
1654 2. If an item is private, it may be accessed by the current module and its
1657 These two cases are surprisingly powerful for creating module hierarchies
1658 exposing public APIs while hiding internal implementation details. To help
1659 explain, here's a few use cases and what they would entail:
1661 * A library developer needs to expose functionality to crates which link
1662 against their library. As a consequence of the first case, this means that
1663 anything which is usable externally must be `pub` from the root down to the
1664 destination item. Any private item in the chain will disallow external
1667 * A crate needs a global available "helper module" to itself, but it doesn't
1668 want to expose the helper module as a public API. To accomplish this, the
1669 root of the crate's hierarchy would have a private module which then
1670 internally has a "public API". Because the entire crate is a descendant of
1671 the root, then the entire local crate can access this private module through
1674 * When writing unit tests for a module, it's often a common idiom to have an
1675 immediate child of the module to-be-tested named `mod test`. This module
1676 could access any items of the parent module through the second case, meaning
1677 that internal implementation details could also be seamlessly tested from the
1680 In the second case, it mentions that a private item "can be accessed" by the
1681 current module and its descendants, but the exact meaning of accessing an item
1682 depends on what the item is. Accessing a module, for example, would mean
1683 looking inside of it (to import more items). On the other hand, accessing a
1684 function would mean that it is invoked. Additionally, path expressions and
1685 import statements are considered to access an item in the sense that the
1686 import/expression is only valid if the destination is in the current visibility
1689 Here's an example of a program which exemplifies the three cases outlined
1693 // This module is private, meaning that no external crate can access this
1694 // module. Because it is private at the root of this current crate, however, any
1695 // module in the crate may access any publicly visible item in this module.
1696 mod crate_helper_module {
1698 // This function can be used by anything in the current crate
1699 pub fn crate_helper() {}
1701 // This function *cannot* be used by anything else in the crate. It is not
1702 // publicly visible outside of the `crate_helper_module`, so only this
1703 // current module and its descendants may access it.
1704 fn implementation_detail() {}
1707 // This function is "public to the root" meaning that it's available to external
1708 // crates linking against this one.
1709 pub fn public_api() {}
1711 // Similarly to 'public_api', this module is public so external crates may look
1714 use crate_helper_module;
1716 pub fn my_method() {
1717 // Any item in the local crate may invoke the helper module's public
1718 // interface through a combination of the two rules above.
1719 crate_helper_module::crate_helper();
1722 // This function is hidden to any module which is not a descendant of
1724 fn my_implementation() {}
1730 fn test_my_implementation() {
1731 // Because this module is a descendant of `submodule`, it's allowed
1732 // to access private items inside of `submodule` without a privacy
1734 super::my_implementation();
1742 For a rust program to pass the privacy checking pass, all paths must be valid
1743 accesses given the two rules above. This includes all use statements,
1744 expressions, types, etc.
1746 ### Re-exporting and Visibility
1748 Rust allows publicly re-exporting items through a `pub use` directive. Because
1749 this is a public directive, this allows the item to be used in the current
1750 module through the rules above. It essentially allows public access into the
1751 re-exported item. For example, this program is valid:
1754 pub use self::implementation::api;
1756 mod implementation {
1765 This means that any external crate referencing `implementation::api::f` would
1766 receive a privacy violation, while the path `api::f` would be allowed.
1768 When re-exporting a private item, it can be thought of as allowing the "privacy
1769 chain" being short-circuited through the reexport instead of passing through
1770 the namespace hierarchy as it normally would.
1774 Any item declaration may have an _attribute_ applied to it. Attributes in Rust
1775 are modeled on Attributes in ECMA-335, with the syntax coming from ECMA-334
1776 (C#). An attribute is a general, free-form metadatum that is interpreted
1777 according to name, convention, and language and compiler version. Attributes
1778 may appear as any of:
1780 * A single identifier, the attribute name
1781 * An identifier followed by the equals sign '=' and a literal, providing a
1783 * An identifier followed by a parenthesized list of sub-attribute arguments
1785 Attributes with a bang ("!") after the hash ("#") apply to the item that the
1786 attribute is declared within. Attributes that do not have a bang after the hash
1787 apply to the item that follows the attribute.
1789 An example of attributes:
1792 // General metadata applied to the enclosing module or crate.
1793 #![crate_type = "lib"]
1795 // A function marked as a unit test
1801 // A conditionally-compiled module
1802 #[cfg(target_os="linux")]
1807 // A lint attribute used to suppress a warning/error
1808 #[allow(non_camel_case_types)]
1812 > **Note:** At some point in the future, the compiler will distinguish between
1813 > language-reserved and user-available attributes. Until then, there is
1814 > effectively no difference between an attribute handled by a loadable syntax
1815 > extension and the compiler.
1817 ### Crate-only attributes
1819 - `crate_name` - specify the crate's crate name.
1820 - `crate_type` - see [linkage](#linkage).
1821 - `feature` - see [compiler features](#compiler-features).
1822 - `no_builtins` - disable optimizing certain code patterns to invocations of
1823 library functions that are assumed to exist
1824 - `no_main` - disable emitting the `main` symbol. Useful when some other
1825 object being linked to defines `main`.
1826 - `no_start` - disable linking to the `native` crate, which specifies the
1827 "start" language item.
1828 - `no_std` - disable linking to the `std` crate.
1829 - `plugin` - load a list of named crates as compiler plugins, e.g.
1830 `#![plugin(foo, bar)]`. Optional arguments for each plugin,
1831 i.e. `#![plugin(foo(... args ...))]`, are provided to the plugin's
1832 registrar function. The `plugin` feature gate is required to use
1834 - `recursion_limit` - Sets the maximum depth for potentially
1835 infinitely-recursive compile-time operations like
1836 auto-dereference or macro expansion. The default is
1837 `#![recursion_limit="64"]`.
1839 ### Module-only attributes
1841 - `no_implicit_prelude` - disable injecting `use std::prelude::*` in this
1843 - `path` - specifies the file to load the module from. `#[path="foo.rs"] mod
1844 bar;` is equivalent to `mod bar { /* contents of foo.rs */ }`. The path is
1845 taken relative to the directory that the current module is in.
1847 ### Function-only attributes
1849 - `main` - indicates that this function should be passed to the entry point,
1850 rather than the function in the crate root named `main`.
1851 - `plugin_registrar` - mark this function as the registration point for
1852 [compiler plugins][plugin], such as loadable syntax extensions.
1853 - `start` - indicates that this function should be used as the entry point,
1854 overriding the "start" language item. See the "start" [language
1855 item](#language-items) for more details.
1856 - `test` - indicates that this function is a test function, to only be compiled
1857 in case of `--test`.
1858 - `should_panic` - indicates that this test function should panic, inverting the success condition.
1859 - `cold` - The function is unlikely to be executed, so optimize it (and calls
1862 ### Static-only attributes
1864 - `thread_local` - on a `static mut`, this signals that the value of this
1865 static may change depending on the current thread. The exact consequences of
1866 this are implementation-defined.
1870 On an `extern` block, the following attributes are interpreted:
1872 - `link_args` - specify arguments to the linker, rather than just the library
1873 name and type. This is feature gated and the exact behavior is
1874 implementation-defined (due to variety of linker invocation syntax).
1875 - `link` - indicate that a native library should be linked to for the
1876 declarations in this block to be linked correctly. `link` supports an optional
1877 `kind` key with three possible values: `dylib`, `static`, and `framework`. See
1878 [external blocks](#external-blocks) for more about external blocks. Two
1879 examples: `#[link(name = "readline")]` and
1880 `#[link(name = "CoreFoundation", kind = "framework")]`.
1881 - `linked_from` - indicates what native library this block of FFI items is
1882 coming from. This attribute is of the form `#[linked_from = "foo"]` where
1883 `foo` is the name of a library in either `#[link]` or a `-l` flag. This
1884 attribute is currently required to export symbols from a Rust dynamic library
1885 on Windows, and it is feature gated behind the `linked_from` feature.
1887 On declarations inside an `extern` block, the following attributes are
1890 - `link_name` - the name of the symbol that this function or static should be
1892 - `linkage` - on a static, this specifies the [linkage
1893 type](http://llvm.org/docs/LangRef.html#linkage-types).
1897 - `repr` - on C-like enums, this sets the underlying type used for
1898 representation. Takes one argument, which is the primitive
1899 type this enum should be represented for, or `C`, which specifies that it
1900 should be the default `enum` size of the C ABI for that platform. Note that
1901 enum representation in C is undefined, and this may be incorrect when the C
1902 code is compiled with certain flags.
1906 - `repr` - specifies the representation to use for this struct. Takes a list
1907 of options. The currently accepted ones are `C` and `packed`, which may be
1908 combined. `C` will use a C ABI compatible struct layout, and `packed` will
1909 remove any padding between fields (note that this is very fragile and may
1910 break platforms which require aligned access).
1912 ### Macro-related attributes
1914 - `macro_use` on a `mod` — macros defined in this module will be visible in the
1915 module's parent, after this module has been included.
1917 - `macro_use` on an `extern crate` — load macros from this crate. An optional
1918 list of names `#[macro_use(foo, bar)]` restricts the import to just those
1919 macros named. The `extern crate` must appear at the crate root, not inside
1920 `mod`, which ensures proper function of the [`$crate` macro
1921 variable](book/macros.html#the-variable-crate).
1923 - `macro_reexport` on an `extern crate` — re-export the named macros.
1925 - `macro_export` - export a macro for cross-crate usage.
1927 - `no_link` on an `extern crate` — even if we load this crate for macros, don't
1928 link it into the output.
1930 See the [macros section of the
1931 book](book/macros.html#scoping-and-macro-importexport) for more information on
1935 ### Miscellaneous attributes
1937 - `export_name` - on statics and functions, this determines the name of the
1939 - `link_section` - on statics and functions, this specifies the section of the
1940 object file that this item's contents will be placed into.
1941 - `no_mangle` - on any item, do not apply the standard name mangling. Set the
1942 symbol for this item to its identifier.
1943 - `simd` - on certain tuple structs, derive the arithmetic operators, which
1944 lower to the target's SIMD instructions, if any; the `simd` feature gate
1945 is necessary to use this attribute.
1946 - `unsafe_destructor_blind_to_params` - on `Drop::drop` method, asserts that the
1947 destructor code (and all potential specializations of that code) will
1948 never attempt to read from nor write to any references with lifetimes
1949 that come in via generic parameters. This is a constraint we cannot
1950 currently express via the type system, and therefore we rely on the
1951 programmer to assert that it holds. Adding this to a Drop impl causes
1952 the associated destructor to be considered "uninteresting" by the
1953 Drop-Check rule, and thus it can help sidestep data ordering
1954 constraints that would otherwise be introduced by the Drop-Check
1955 rule. Such sidestepping of the constraints, if done incorrectly, can
1956 lead to undefined behavior (in the form of reading or writing to data
1957 outside of its dynamic extent), and thus this attribute has the word
1958 "unsafe" in its name. To use this, the
1959 `unsafe_destructor_blind_to_params` feature gate must be enabled.
1960 - `unsafe_no_drop_flag` - on structs, remove the flag that prevents
1961 destructors from being run twice. Destructors might be run multiple times on
1962 the same object with this attribute. To use this, the `unsafe_no_drop_flag` feature
1963 gate must be enabled.
1964 - `doc` - Doc comments such as `/// foo` are equivalent to `#[doc = "foo"]`.
1965 - `rustc_on_unimplemented` - Write a custom note to be shown along with the error
1966 when the trait is found to be unimplemented on a type.
1967 You may use format arguments like `{T}`, `{A}` to correspond to the
1968 types at the point of use corresponding to the type parameters of the
1969 trait of the same name. `{Self}` will be replaced with the type that is supposed
1970 to implement the trait but doesn't. To use this, the `on_unimplemented` feature gate
1973 ### Conditional compilation
1975 Sometimes one wants to have different compiler outputs from the same code,
1976 depending on build target, such as targeted operating system, or to enable
1979 There are two kinds of configuration options, one that is either defined or not
1980 (`#[cfg(foo)]`), and the other that contains a string that can be checked
1981 against (`#[cfg(bar = "baz")]`). Currently, only compiler-defined configuration
1982 options can have the latter form.
1985 // The function is only included in the build when compiling for OSX
1986 #[cfg(target_os = "macos")]
1991 // This function is only included when either foo or bar is defined
1992 #[cfg(any(foo, bar))]
1993 fn needs_foo_or_bar() {
1997 // This function is only included when compiling for a unixish OS with a 32-bit
1999 #[cfg(all(unix, target_pointer_width = "32"))]
2000 fn on_32bit_unix() {
2004 // This function is only included when foo is not defined
2006 fn needs_not_foo() {
2011 This illustrates some conditional compilation can be achieved using the
2012 `#[cfg(...)]` attribute. `any`, `all` and `not` can be used to assemble
2013 arbitrarily complex configurations through nesting.
2015 The following configurations must be defined by the implementation:
2017 * `debug_assertions` - Enabled by default when compiling without optimizations.
2018 This can be used to enable extra debugging code in development but not in
2019 production. For example, it controls the behavior of the standard library's
2020 `debug_assert!` macro.
2021 * `target_arch = "..."` - Target CPU architecture, such as `"x86"`, `"x86_64"`
2022 `"mips"`, `"powerpc"`, `"arm"`, or `"aarch64"`.
2023 * `target_endian = "..."` - Endianness of the target CPU, either `"little"` or
2025 * `target_env = ".."` - An option provided by the compiler by default
2026 describing the runtime environment of the target platform. Some examples of
2027 this are `musl` for builds targeting the MUSL libc implementation, `msvc` for
2028 Windows builds targeting MSVC, and `gnu` frequently the rest of the time. This
2029 option may also be blank on some platforms.
2030 * `target_family = "..."` - Operating system family of the target, e. g.
2031 `"unix"` or `"windows"`. The value of this configuration option is defined
2032 as a configuration itself, like `unix` or `windows`.
2033 * `target_os = "..."` - Operating system of the target, examples include
2034 `"windows"`, `"macos"`, `"ios"`, `"linux"`, `"android"`, `"freebsd"`, `"dragonfly"`,
2035 `"bitrig"` , `"openbsd"` or `"netbsd"`.
2036 * `target_pointer_width = "..."` - Target pointer width in bits. This is set
2037 to `"32"` for targets with 32-bit pointers, and likewise set to `"64"` for
2039 * `target_vendor = "..."` - Vendor of the target, for example `apple`, `pc`, or
2041 * `test` - Enabled when compiling the test harness (using the `--test` flag).
2042 * `unix` - See `target_family`.
2043 * `windows` - See `target_family`.
2045 You can also set another attribute based on a `cfg` variable with `cfg_attr`:
2051 Will be the same as `#[b]` if `a` is set by `cfg`, and nothing otherwise.
2053 ### Lint check attributes
2055 A lint check names a potentially undesirable coding pattern, such as
2056 unreachable code or omitted documentation, for the static entity to which the
2059 For any lint check `C`:
2061 * `allow(C)` overrides the check for `C` so that violations will go
2063 * `deny(C)` signals an error after encountering a violation of `C`,
2064 * `forbid(C)` is the same as `deny(C)`, but also forbids changing the lint
2066 * `warn(C)` warns about violations of `C` but continues compilation.
2068 The lint checks supported by the compiler can be found via `rustc -W help`,
2069 along with their default settings. [Compiler
2070 plugins](book/compiler-plugins.html#lint-plugins) can provide additional lint checks.
2074 // Missing documentation is ignored here
2075 #[allow(missing_docs)]
2076 pub fn undocumented_one() -> i32 { 1 }
2078 // Missing documentation signals a warning here
2079 #[warn(missing_docs)]
2080 pub fn undocumented_too() -> i32 { 2 }
2082 // Missing documentation signals an error here
2083 #[deny(missing_docs)]
2084 pub fn undocumented_end() -> i32 { 3 }
2088 This example shows how one can use `allow` and `warn` to toggle a particular
2092 #[warn(missing_docs)]
2094 #[allow(missing_docs)]
2096 // Missing documentation is ignored here
2097 pub fn undocumented_one() -> i32 { 1 }
2099 // Missing documentation signals a warning here,
2100 // despite the allow above.
2101 #[warn(missing_docs)]
2102 pub fn undocumented_two() -> i32 { 2 }
2105 // Missing documentation signals a warning here
2106 pub fn undocumented_too() -> i32 { 3 }
2110 This example shows how one can use `forbid` to disallow uses of `allow` for
2114 #[forbid(missing_docs)]
2116 // Attempting to toggle warning signals an error here
2117 #[allow(missing_docs)]
2119 pub fn undocumented_too() -> i32 { 2 }
2125 Some primitive Rust operations are defined in Rust code, rather than being
2126 implemented directly in C or assembly language. The definitions of these
2127 operations have to be easy for the compiler to find. The `lang` attribute
2128 makes it possible to declare these operations. For example, the `str` module
2129 in the Rust standard library defines the string equality function:
2133 pub fn eq_slice(a: &str, b: &str) -> bool {
2138 The name `str_eq` has a special meaning to the Rust compiler, and the presence
2139 of this definition means that it will use this definition when generating calls
2140 to the string equality function.
2142 The set of language items is currently considered unstable. A complete
2143 list of the built-in language items will be added in the future.
2145 ### Inline attributes
2147 The inline attribute suggests that the compiler should place a copy of
2148 the function or static in the caller, rather than generating code to
2149 call the function or access the static where it is defined.
2151 The compiler automatically inlines functions based on internal heuristics.
2152 Incorrectly inlining functions can actually make the program slower, so it
2153 should be used with care.
2155 `#[inline]` and `#[inline(always)]` always cause the function to be serialized
2156 into the crate metadata to allow cross-crate inlining.
2158 There are three different types of inline attributes:
2160 * `#[inline]` hints the compiler to perform an inline expansion.
2161 * `#[inline(always)]` asks the compiler to always perform an inline expansion.
2162 * `#[inline(never)]` asks the compiler to never perform an inline expansion.
2166 The `derive` attribute allows certain traits to be automatically implemented
2167 for data structures. For example, the following will create an `impl` for the
2168 `PartialEq` and `Clone` traits for `Foo`, the type parameter `T` will be given
2169 the `PartialEq` or `Clone` constraints for the appropriate `impl`:
2172 #[derive(PartialEq, Clone)]
2179 The generated `impl` for `PartialEq` is equivalent to
2182 # struct Foo<T> { a: i32, b: T }
2183 impl<T: PartialEq> PartialEq for Foo<T> {
2184 fn eq(&self, other: &Foo<T>) -> bool {
2185 self.a == other.a && self.b == other.b
2188 fn ne(&self, other: &Foo<T>) -> bool {
2189 self.a != other.a || self.b != other.b
2194 ### Compiler Features
2196 Certain aspects of Rust may be implemented in the compiler, but they're not
2197 necessarily ready for every-day use. These features are often of "prototype
2198 quality" or "almost production ready", but may not be stable enough to be
2199 considered a full-fledged language feature.
2201 For this reason, Rust recognizes a special crate-level attribute of the form:
2204 #![feature(feature1, feature2, feature3)]
2207 This directive informs the compiler that the feature list: `feature1`,
2208 `feature2`, and `feature3` should all be enabled. This is only recognized at a
2209 crate-level, not at a module-level. Without this directive, all features are
2210 considered off, and using the features will result in a compiler error.
2212 The currently implemented features of the reference compiler are:
2214 * `advanced_slice_patterns` - See the [match expressions](#match-expressions)
2215 section for discussion; the exact semantics of
2216 slice patterns are subject to change, so some types
2219 * `slice_patterns` - OK, actually, slice patterns are just scary and
2220 completely unstable.
2222 * `asm` - The `asm!` macro provides a means for inline assembly. This is often
2223 useful, but the exact syntax for this feature along with its
2224 semantics are likely to change, so this macro usage must be opted
2227 * `associated_consts` - Allows constants to be defined in `impl` and `trait`
2228 blocks, so that they can be associated with a type or
2229 trait in a similar manner to methods and associated
2232 * `box_patterns` - Allows `box` patterns, the exact semantics of which
2233 is subject to change.
2235 * `box_syntax` - Allows use of `box` expressions, the exact semantics of which
2236 is subject to change.
2238 * `cfg_target_vendor` - Allows conditional compilation using the `target_vendor`
2239 matcher which is subject to change.
2241 * `concat_idents` - Allows use of the `concat_idents` macro, which is in many
2242 ways insufficient for concatenating identifiers, and may be
2243 removed entirely for something more wholesome.
2245 * `custom_attribute` - Allows the usage of attributes unknown to the compiler
2246 so that new attributes can be added in a backwards compatible
2249 * `custom_derive` - Allows the use of `#[derive(Foo,Bar)]` as sugar for
2250 `#[derive_Foo] #[derive_Bar]`, which can be user-defined syntax
2253 * `intrinsics` - Allows use of the "rust-intrinsics" ABI. Compiler intrinsics
2254 are inherently unstable and no promise about them is made.
2256 * `lang_items` - Allows use of the `#[lang]` attribute. Like `intrinsics`,
2257 lang items are inherently unstable and no promise about them
2260 * `link_args` - This attribute is used to specify custom flags to the linker,
2261 but usage is strongly discouraged. The compiler's usage of the
2262 system linker is not guaranteed to continue in the future, and
2263 if the system linker is not used then specifying custom flags
2264 doesn't have much meaning.
2266 * `link_llvm_intrinsics` – Allows linking to LLVM intrinsics via
2267 `#[link_name="llvm.*"]`.
2269 * `linkage` - Allows use of the `linkage` attribute, which is not portable.
2271 * `log_syntax` - Allows use of the `log_syntax` macro attribute, which is a
2272 nasty hack that will certainly be removed.
2274 * `main` - Allows use of the `#[main]` attribute, which changes the entry point
2275 into a Rust program. This capability is subject to change.
2277 * `macro_reexport` - Allows macros to be re-exported from one crate after being imported
2278 from another. This feature was originally designed with the sole
2279 use case of the Rust standard library in mind, and is subject to
2282 * `non_ascii_idents` - The compiler supports the use of non-ascii identifiers,
2283 but the implementation is a little rough around the
2284 edges, so this can be seen as an experimental feature
2285 for now until the specification of identifiers is fully
2288 * `no_std` - Allows the `#![no_std]` crate attribute, which disables the implicit
2289 `extern crate std`. This typically requires use of the unstable APIs
2290 behind the libstd "facade", such as libcore and libcollections. It
2291 may also cause problems when using syntax extensions, including
2294 * `on_unimplemented` - Allows the `#[rustc_on_unimplemented]` attribute, which allows
2295 trait definitions to add specialized notes to error messages
2296 when an implementation was expected but not found.
2298 * `optin_builtin_traits` - Allows the definition of default and negative trait
2299 implementations. Experimental.
2301 * `plugin` - Usage of [compiler plugins][plugin] for custom lints or syntax extensions.
2302 These depend on compiler internals and are subject to change.
2304 * `plugin_registrar` - Indicates that a crate provides [compiler plugins][plugin].
2306 * `quote` - Allows use of the `quote_*!` family of macros, which are
2307 implemented very poorly and will likely change significantly
2308 with a proper implementation.
2310 * `rustc_attrs` - Gates internal `#[rustc_*]` attributes which may be
2311 for internal use only or have meaning added to them in the future.
2313 * `rustc_diagnostic_macros`- A mysterious feature, used in the implementation
2314 of rustc, not meant for mortals.
2316 * `simd` - Allows use of the `#[simd]` attribute, which is overly simple and
2317 not the SIMD interface we want to expose in the long term.
2319 * `simd_ffi` - Allows use of SIMD vectors in signatures for foreign functions.
2320 The SIMD interface is subject to change.
2322 * `staged_api` - Allows usage of stability markers and `#![staged_api]` in a
2323 crate. Stability markers are also attributes: `#[stable]`,
2324 `#[unstable]`, and `#[deprecated]` are the three levels.
2326 * `start` - Allows use of the `#[start]` attribute, which changes the entry point
2327 into a Rust program. This capability, especially the signature for the
2328 annotated function, is subject to change.
2330 * `struct_variant` - Structural enum variants (those with named fields). It is
2331 currently unknown whether this style of enum variant is as
2332 fully supported as the tuple-forms, and it's not certain
2333 that this style of variant should remain in the language.
2334 For now this style of variant is hidden behind a feature
2337 * `thread_local` - The usage of the `#[thread_local]` attribute is experimental
2338 and should be seen as unstable. This attribute is used to
2339 declare a `static` as being unique per-thread leveraging
2340 LLVM's implementation which works in concert with the kernel
2341 loader and dynamic linker. This is not necessarily available
2342 on all platforms, and usage of it is discouraged.
2344 * `trace_macros` - Allows use of the `trace_macros` macro, which is a nasty
2345 hack that will certainly be removed.
2347 * `unboxed_closures` - Rust's new closure design, which is currently a work in
2348 progress feature with many known bugs.
2350 * `unsafe_no_drop_flag` - Allows use of the `#[unsafe_no_drop_flag]` attribute,
2351 which removes hidden flag added to a type that
2352 implements the `Drop` trait. The design for the
2353 `Drop` flag is subject to change, and this feature
2354 may be removed in the future.
2356 * `unmarked_api` - Allows use of items within a `#![staged_api]` crate
2357 which have not been marked with a stability marker.
2358 Such items should not be allowed by the compiler to exist,
2359 so if you need this there probably is a compiler bug.
2361 * `visible_private_types` - Allows public APIs to expose otherwise private
2362 types, e.g. as the return type of a public function.
2363 This capability may be removed in the future.
2365 * `allow_internal_unstable` - Allows `macro_rules!` macros to be tagged with the
2366 `#[allow_internal_unstable]` attribute, designed
2367 to allow `std` macros to call
2368 `#[unstable]`/feature-gated functionality
2369 internally without imposing on callers
2370 (i.e. making them behave like function calls in
2371 terms of encapsulation).
2372 * - `default_type_parameter_fallback` - Allows type parameter defaults to
2373 influence type inference.
2374 * - `braced_empty_structs` - Allows use of empty structs and enum variants with braces.
2376 If a feature is promoted to a language feature, then all existing programs will
2377 start to receive compilation warnings about `#![feature]` directives which enabled
2378 the new feature (because the directive is no longer necessary). However, if a
2379 feature is decided to be removed from the language, errors will be issued (if
2380 there isn't a parser error first). The directive in this case is no longer
2381 necessary, and it's likely that existing code will break if the feature isn't
2384 If an unknown feature is found in a directive, it results in a compiler error.
2385 An unknown feature is one which has never been recognized by the compiler.
2387 # Statements and expressions
2389 Rust is _primarily_ an expression language. This means that most forms of
2390 value-producing or effect-causing evaluation are directed by the uniform syntax
2391 category of _expressions_. Each kind of expression can typically _nest_ within
2392 each other kind of expression, and rules for evaluation of expressions involve
2393 specifying both the value produced by the expression and the order in which its
2394 sub-expressions are themselves evaluated.
2396 In contrast, statements in Rust serve _mostly_ to contain and explicitly
2397 sequence expression evaluation.
2401 A _statement_ is a component of a block, which is in turn a component of an
2402 outer [expression](#expressions) or [function](#functions).
2404 Rust has two kinds of statement: [declaration
2405 statements](#declaration-statements) and [expression
2406 statements](#expression-statements).
2408 ### Declaration statements
2410 A _declaration statement_ is one that introduces one or more *names* into the
2411 enclosing statement block. The declared names may denote new variables or new
2414 #### Item declarations
2416 An _item declaration statement_ has a syntactic form identical to an
2417 [item](#items) declaration within a module. Declaring an item — a
2418 function, enumeration, struct, type, static, trait, implementation or module
2419 — locally within a statement block is simply a way of restricting its
2420 scope to a narrow region containing all of its uses; it is otherwise identical
2421 in meaning to declaring the item outside the statement block.
2423 > **Note**: there is no implicit capture of the function's dynamic environment when
2424 > declaring a function-local item.
2426 #### Variable declarations
2428 A _variable declaration_ introduces a new set of variable, given by a pattern. The
2429 pattern may be followed by a type annotation, and/or an initializer expression.
2430 When no type annotation is given, the compiler will infer the type, or signal
2431 an error if insufficient type information is available for definite inference.
2432 Any variables introduced by a variable declaration are visible from the point of
2433 declaration until the end of the enclosing block scope.
2435 ### Expression statements
2437 An _expression statement_ is one that evaluates an [expression](#expressions)
2438 and ignores its result. The type of an expression statement `e;` is always
2439 `()`, regardless of the type of `e`. As a rule, an expression statement's
2440 purpose is to trigger the effects of evaluating its expression.
2444 An expression may have two roles: it always produces a *value*, and it may have
2445 *effects* (otherwise known as "side effects"). An expression *evaluates to* a
2446 value, and has effects during *evaluation*. Many expressions contain
2447 sub-expressions (operands). The meaning of each kind of expression dictates
2450 * Whether or not to evaluate the sub-expressions when evaluating the expression
2451 * The order in which to evaluate the sub-expressions
2452 * How to combine the sub-expressions' values to obtain the value of the expression
2454 In this way, the structure of expressions dictates the structure of execution.
2455 Blocks are just another kind of expression, so blocks, statements, expressions,
2456 and blocks again can recursively nest inside each other to an arbitrary depth.
2458 #### Lvalues, rvalues and temporaries
2460 Expressions are divided into two main categories: _lvalues_ and _rvalues_.
2461 Likewise within each expression, sub-expressions may occur in _lvalue context_
2462 or _rvalue context_. The evaluation of an expression depends both on its own
2463 category and the context it occurs within.
2465 An lvalue is an expression that represents a memory location. These expressions
2466 are [paths](#path-expressions) (which refer to local variables, function and
2467 method arguments, or static variables), dereferences (`*expr`), [indexing
2468 expressions](#index-expressions) (`expr[expr]`), and [field
2469 references](#field-expressions) (`expr.f`). All other expressions are rvalues.
2471 The left operand of an [assignment](#assignment-expressions) or
2472 [compound-assignment](#compound-assignment-expressions) expression is
2473 an lvalue context, as is the single operand of a unary
2474 [borrow](#unary-operator-expressions). The discriminant or subject of
2475 a [match expression](#match-expressions) may be an lvalue context, if
2476 ref bindings are made, but is otherwise an rvalue context. All other
2477 expression contexts are rvalue contexts.
2479 When an lvalue is evaluated in an _lvalue context_, it denotes a memory
2480 location; when evaluated in an _rvalue context_, it denotes the value held _in_
2481 that memory location.
2483 ##### Temporary lifetimes
2485 When an rvalue is used in an lvalue context, a temporary un-named
2486 lvalue is created and used instead. The lifetime of temporary values
2487 is typically the innermost enclosing statement; the tail expression of
2488 a block is considered part of the statement that encloses the block.
2490 When a temporary rvalue is being created that is assigned into a `let`
2491 declaration, however, the temporary is created with the lifetime of
2492 the enclosing block instead, as using the enclosing statement (the
2493 `let` declaration) would be a guaranteed error (since a pointer to the
2494 temporary would be stored into a variable, but the temporary would be
2495 freed before the variable could be used). The compiler uses simple
2496 syntactic rules to decide which values are being assigned into a `let`
2497 binding, and therefore deserve a longer temporary lifetime.
2499 Here are some examples:
2501 - `let x = foo(&temp())`. The expression `temp()` is an rvalue. As it
2502 is being borrowed, a temporary is created which will be freed after
2503 the innermost enclosing statement (the `let` declaration, in this case).
2504 - `let x = temp().foo()`. This is the same as the previous example,
2505 except that the value of `temp()` is being borrowed via autoref on a
2506 method-call. Here we are assuming that `foo()` is an `&self` method
2507 defined in some trait, say `Foo`. In other words, the expression
2508 `temp().foo()` is equivalent to `Foo::foo(&temp())`.
2509 - `let x = &temp()`. Here, the same temporary is being assigned into
2510 `x`, rather than being passed as a parameter, and hence the
2511 temporary's lifetime is considered to be the enclosing block.
2512 - `let x = SomeStruct { foo: &temp() }`. As in the previous case, the
2513 temporary is assigned into a struct which is then assigned into a
2514 binding, and hence it is given the lifetime of the enclosing block.
2515 - `let x = [ &temp() ]`. As in the previous case, the
2516 temporary is assigned into an array which is then assigned into a
2517 binding, and hence it is given the lifetime of the enclosing block.
2518 - `let ref x = temp()`. In this case, the temporary is created using a ref binding,
2519 but the result is the same: the lifetime is extended to the enclosing block.
2521 #### Moved and copied types
2523 When a [local variable](#variables) is used as an
2524 [rvalue](#lvalues-rvalues-and-temporaries), the variable will be copied
2525 if its type implements `Copy`. All others are moved.
2527 ### Literal expressions
2529 A _literal expression_ consists of one of the [literal](#literals) forms
2530 described earlier. It directly describes a number, character, string, boolean
2531 value, or the unit value.
2535 "hello"; // string type
2536 '5'; // character type
2540 ### Path expressions
2542 A [path](#paths) used as an expression context denotes either a local variable
2543 or an item. Path expressions are [lvalues](#lvalues-rvalues-and-temporaries).
2545 ### Tuple expressions
2547 Tuples are written by enclosing zero or more comma-separated expressions in
2548 parentheses. They are used to create [tuple-typed](#tuple-types) values.
2552 ("a", 4usize, true);
2555 You can disambiguate a single-element tuple from a value in parentheses with a
2559 (0,); // single-element tuple
2560 (0); // zero in parentheses
2563 ### Struct expressions
2565 There are several forms of struct expressions. A _struct expression_
2566 consists of the [path](#paths) of a [struct item](#structs), followed by
2567 a brace-enclosed list of one or more comma-separated name-value pairs,
2568 providing the field values of a new instance of the struct. A field name
2569 can be any identifier, and is separated from its value expression by a colon.
2570 The location denoted by a struct field is mutable if and only if the
2571 enclosing struct is mutable.
2573 A _tuple struct expression_ consists of the [path](#paths) of a [struct
2574 item](#structs), followed by a parenthesized list of one or more
2575 comma-separated expressions (in other words, the path of a struct item
2576 followed by a tuple expression). The struct item must be a tuple struct
2579 A _unit-like struct expression_ consists only of the [path](#paths) of a
2580 [struct item](#structs).
2582 The following are examples of struct expressions:
2585 # struct Point { x: f64, y: f64 }
2586 # struct TuplePoint(f64, f64);
2587 # mod game { pub struct User<'a> { pub name: &'a str, pub age: u32, pub score: usize } }
2588 # struct Cookie; fn some_fn<T>(t: T) {}
2589 Point {x: 10.0, y: 20.0};
2590 TuplePoint(10.0, 20.0);
2591 let u = game::User {name: "Joe", age: 35, score: 100_000};
2592 some_fn::<Cookie>(Cookie);
2595 A struct expression forms a new value of the named struct type. Note
2596 that for a given *unit-like* struct type, this will always be the same
2599 A struct expression can terminate with the syntax `..` followed by an
2600 expression to denote a functional update. The expression following `..` (the
2601 base) must have the same struct type as the new struct type being formed.
2602 The entire expression denotes the result of constructing a new struct (with
2603 the same type as the base expression) with the given values for the fields that
2604 were explicitly specified and the values in the base expression for all other
2608 # struct Point3d { x: i32, y: i32, z: i32 }
2609 let base = Point3d {x: 1, y: 2, z: 3};
2610 Point3d {y: 0, z: 10, .. base};
2613 ### Block expressions
2615 A _block expression_ is similar to a module in terms of the declarations that
2616 are possible. Each block conceptually introduces a new namespace scope. Use
2617 items can bring new names into scopes and declared items are in scope for only
2620 A block will execute each statement sequentially, and then execute the
2621 expression (if given). If the block ends in a statement, its value is `()`:
2624 let x: () = { println!("Hello."); };
2627 If it ends in an expression, its value and type are that of the expression:
2630 let x: i32 = { println!("Hello."); 5 };
2635 ### Method-call expressions
2637 A _method call_ consists of an expression followed by a single dot, an
2638 identifier, and a parenthesized expression-list. Method calls are resolved to
2639 methods on specific traits, either statically dispatching to a method if the
2640 exact `self`-type of the left-hand-side is known, or dynamically dispatching if
2641 the left-hand-side expression is an indirect [trait object](#trait-objects).
2643 ### Field expressions
2645 A _field expression_ consists of an expression followed by a single dot and an
2646 identifier, when not immediately followed by a parenthesized expression-list
2647 (the latter is a [method call expression](#method-call-expressions)). A field
2648 expression denotes a field of a [struct](#struct-types).
2653 (Struct {a: 10, b: 20}).a;
2656 A field access is an [lvalue](#lvalues-rvalues-and-temporaries) referring to
2657 the value of that field. When the type providing the field inherits mutability,
2658 it can be [assigned](#assignment-expressions) to.
2660 Also, if the type of the expression to the left of the dot is a
2661 pointer, it is automatically dereferenced as many times as necessary
2662 to make the field access possible. In cases of ambiguity, we prefer
2663 fewer autoderefs to more.
2665 ### Array expressions
2667 An [array](#array-and-slice-types) _expression_ is written by enclosing zero
2668 or more comma-separated expressions of uniform type in square brackets.
2670 In the `[expr ';' expr]` form, the expression after the `';'` must be a
2671 constant expression that can be evaluated at compile time, such as a
2672 [literal](#literals) or a [static item](#static-items).
2676 ["a", "b", "c", "d"];
2677 [0; 128]; // array with 128 zeros
2678 [0u8, 0u8, 0u8, 0u8];
2681 ### Index expressions
2683 [Array](#array-and-slice-types)-typed expressions can be indexed by
2684 writing a square-bracket-enclosed expression (the index) after them. When the
2685 array is mutable, the resulting [lvalue](#lvalues-rvalues-and-temporaries) can
2688 Indices are zero-based, and may be of any integral type. Vector access is
2689 bounds-checked at compile-time for constant arrays being accessed with a constant index value.
2690 Otherwise a check will be performed at run-time that will put the thread in a _panicked state_ if it fails.
2695 let x = (["a", "b"])[10]; // compiler error: const index-expr is out of bounds
2698 let y = (["a", "b"])[n]; // panics
2700 let arr = ["a", "b"];
2704 Also, if the type of the expression to the left of the brackets is a
2705 pointer, it is automatically dereferenced as many times as necessary
2706 to make the indexing possible. In cases of ambiguity, we prefer fewer
2709 ### Range expressions
2711 The `..` operator will construct an object of one of the `std::ops::Range` variants.
2714 1..2; // std::ops::Range
2715 3..; // std::ops::RangeFrom
2716 ..4; // std::ops::RangeTo
2717 ..; // std::ops::RangeFull
2720 The following expressions are equivalent.
2723 let x = std::ops::Range {start: 0, end: 10};
2729 ### Unary operator expressions
2731 Rust defines the following unary operators. They are all written as prefix operators,
2732 before the expression they apply to.
2735 : Negation. May only be applied to numeric types.
2737 : Dereference. When applied to a [pointer](#pointer-types) it denotes the
2738 pointed-to location. For pointers to mutable locations, the resulting
2739 [lvalue](#lvalues-rvalues-and-temporaries) can be assigned to.
2740 On non-pointer types, it calls the `deref` method of the `std::ops::Deref`
2741 trait, or the `deref_mut` method of the `std::ops::DerefMut` trait (if
2742 implemented by the type and required for an outer expression that will or
2743 could mutate the dereference), and produces the result of dereferencing the
2744 `&` or `&mut` borrowed pointer returned from the overload method.
2746 : Logical negation. On the boolean type, this flips between `true` and
2747 `false`. On integer types, this inverts the individual bits in the
2748 two's complement representation of the value.
2750 : Borrowing. When applied to an lvalue, these operators produce a
2751 reference (pointer) to the lvalue. The lvalue is also placed into
2752 a borrowed state for the duration of the reference. For a shared
2753 borrow (`&`), this implies that the lvalue may not be mutated, but
2754 it may be read or shared again. For a mutable borrow (`&mut`), the
2755 lvalue may not be accessed in any way until the borrow expires.
2756 If the `&` or `&mut` operators are applied to an rvalue, a
2757 temporary value is created; the lifetime of this temporary value
2758 is defined by [syntactic rules](#temporary-lifetimes).
2760 ### Binary operator expressions
2762 Binary operators expressions are given in terms of [operator
2763 precedence](#operator-precedence).
2765 #### Arithmetic operators
2767 Binary arithmetic expressions are syntactic sugar for calls to built-in traits,
2768 defined in the `std::ops` module of the `std` library. This means that
2769 arithmetic operators can be overridden for user-defined types. The default
2770 meaning of the operators on standard types is given here.
2773 : Addition and array/string concatenation.
2774 Calls the `add` method on the `std::ops::Add` trait.
2777 Calls the `sub` method on the `std::ops::Sub` trait.
2780 Calls the `mul` method on the `std::ops::Mul` trait.
2783 Calls the `div` method on the `std::ops::Div` trait.
2786 Calls the `rem` method on the `std::ops::Rem` trait.
2788 #### Bitwise operators
2790 Like the [arithmetic operators](#arithmetic-operators), bitwise operators are
2791 syntactic sugar for calls to methods of built-in traits. This means that
2792 bitwise operators can be overridden for user-defined types. The default
2793 meaning of the operators on standard types is given here. Bitwise `&`, `|` and
2794 `^` applied to boolean arguments are equivalent to logical `&&`, `||` and `!=`
2795 evaluated in non-lazy fashion.
2799 Calls the `bitand` method of the `std::ops::BitAnd` trait.
2801 : Bitwise inclusive OR.
2802 Calls the `bitor` method of the `std::ops::BitOr` trait.
2804 : Bitwise exclusive OR.
2805 Calls the `bitxor` method of the `std::ops::BitXor` trait.
2808 Calls the `shl` method of the `std::ops::Shl` trait.
2810 : Right shift (arithmetic).
2811 Calls the `shr` method of the `std::ops::Shr` trait.
2813 #### Lazy boolean operators
2815 The operators `||` and `&&` may be applied to operands of boolean type. The
2816 `||` operator denotes logical 'or', and the `&&` operator denotes logical
2817 'and'. They differ from `|` and `&` in that the right-hand operand is only
2818 evaluated when the left-hand operand does not already determine the result of
2819 the expression. That is, `||` only evaluates its right-hand operand when the
2820 left-hand operand evaluates to `false`, and `&&` only when it evaluates to
2823 #### Comparison operators
2825 Comparison operators are, like the [arithmetic
2826 operators](#arithmetic-operators), and [bitwise operators](#bitwise-operators),
2827 syntactic sugar for calls to built-in traits. This means that comparison
2828 operators can be overridden for user-defined types. The default meaning of the
2829 operators on standard types is given here.
2833 Calls the `eq` method on the `std::cmp::PartialEq` trait.
2836 Calls the `ne` method on the `std::cmp::PartialEq` trait.
2839 Calls the `lt` method on the `std::cmp::PartialOrd` trait.
2842 Calls the `gt` method on the `std::cmp::PartialOrd` trait.
2844 : Less than or equal.
2845 Calls the `le` method on the `std::cmp::PartialOrd` trait.
2847 : Greater than or equal.
2848 Calls the `ge` method on the `std::cmp::PartialOrd` trait.
2850 #### Type cast expressions
2852 A type cast expression is denoted with the binary operator `as`.
2854 Executing an `as` expression casts the value on the left-hand side to the type
2855 on the right-hand side.
2857 An example of an `as` expression:
2860 # fn sum(values: &[f64]) -> f64 { 0.0 }
2861 # fn len(values: &[f64]) -> i32 { 0 }
2863 fn average(values: &[f64]) -> f64 {
2864 let sum: f64 = sum(values);
2865 let size: f64 = len(values) as f64;
2870 Some of the conversions which can be done through the `as` operator
2871 can also be done implicitly at various points in the program, such as
2872 argument passing and assignment to a `let` binding with an explicit
2873 type. Implicit conversions are limited to "harmless" conversions that
2874 do not lose information and which have minimal or no risk of
2875 surprising side-effects on the dynamic execution semantics.
2877 #### Assignment expressions
2879 An _assignment expression_ consists of an
2880 [lvalue](#lvalues-rvalues-and-temporaries) expression followed by an equals
2881 sign (`=`) and an [rvalue](#lvalues-rvalues-and-temporaries) expression.
2883 Evaluating an assignment expression [either copies or
2884 moves](#moved-and-copied-types) its right-hand operand to its left-hand
2893 #### Compound assignment expressions
2895 The `+`, `-`, `*`, `/`, `%`, `&`, `|`, `^`, `<<`, and `>>` operators may be
2896 composed with the `=` operator. The expression `lval OP= val` is equivalent to
2897 `lval = lval OP val`. For example, `x = x + 1` may be written as `x += 1`.
2899 Any such expression always has the [`unit`](#tuple-types) type.
2901 #### Operator precedence
2903 The precedence of Rust binary operators is ordered as follows, going from
2906 ```{.text .precedence}
2920 Operators at the same precedence level are evaluated left-to-right. [Unary
2921 operators](#unary-operator-expressions) have the same precedence level and are
2922 stronger than any of the binary operators.
2924 ### Grouped expressions
2926 An expression enclosed in parentheses evaluates to the result of the enclosed
2927 expression. Parentheses can be used to explicitly specify evaluation order
2928 within an expression.
2930 An example of a parenthesized expression:
2933 let x: i32 = (2 + 3) * 4;
2937 ### Call expressions
2939 A _call expression_ invokes a function, providing zero or more input variables
2940 and an optional location to move the function's output into. If the function
2941 eventually returns, then the expression completes.
2943 Some examples of call expressions:
2946 # fn add(x: i32, y: i32) -> i32 { 0 }
2948 let x: i32 = add(1i32, 2i32);
2949 let pi: Result<f32, _> = "3.14".parse();
2952 ### Lambda expressions
2954 A _lambda expression_ (sometimes called an "anonymous function expression")
2955 defines a function and denotes it as a value, in a single expression. A lambda
2956 expression is a pipe-symbol-delimited (`|`) list of identifiers followed by an
2959 A lambda expression denotes a function that maps a list of parameters
2960 (`ident_list`) onto the expression that follows the `ident_list`. The
2961 identifiers in the `ident_list` are the parameters to the function. These
2962 parameters' types need not be specified, as the compiler infers them from
2965 Lambda expressions are most useful when passing functions as arguments to other
2966 functions, as an abbreviation for defining and capturing a separate function.
2968 Significantly, lambda expressions _capture their environment_, which regular
2969 [function definitions](#functions) do not. The exact type of capture depends
2970 on the [function type](#function-types) inferred for the lambda expression. In
2971 the simplest and least-expensive form (analogous to a ```|| { }``` expression),
2972 the lambda expression captures its environment by reference, effectively
2973 borrowing pointers to all outer variables mentioned inside the function.
2974 Alternately, the compiler may infer that a lambda expression should copy or
2975 move values (depending on their type) from the environment into the lambda
2976 expression's captured environment.
2978 In this example, we define a function `ten_times` that takes a higher-order
2979 function argument, and we then call it with a lambda expression as an argument:
2982 fn ten_times<F>(f: F) where F: Fn(i32) {
2983 for index in 0..10 {
2988 ten_times(|j| println!("hello, {}", j));
2993 A `loop` expression denotes an infinite loop.
2995 A `loop` expression may optionally have a _label_. The label is written as
2996 a lifetime preceding the loop expression, as in `'foo: loop{ }`. If a
2997 label is present, then labeled `break` and `continue` expressions nested
2998 within this loop may exit out of this loop or return control to its head.
2999 See [break expressions](#break-expressions) and [continue
3000 expressions](#continue-expressions).
3002 ### `break` expressions
3004 A `break` expression has an optional _label_. If the label is absent, then
3005 executing a `break` expression immediately terminates the innermost loop
3006 enclosing it. It is only permitted in the body of a loop. If the label is
3007 present, then `break 'foo` terminates the loop with label `'foo`, which need not
3008 be the innermost label enclosing the `break` expression, but must enclose it.
3010 ### `continue` expressions
3012 A `continue` expression has an optional _label_. If the label is absent, then
3013 executing a `continue` expression immediately terminates the current iteration
3014 of the innermost loop enclosing it, returning control to the loop *head*. In
3015 the case of a `while` loop, the head is the conditional expression controlling
3016 the loop. In the case of a `for` loop, the head is the call-expression
3017 controlling the loop. If the label is present, then `continue 'foo` returns
3018 control to the head of the loop with label `'foo`, which need not be the
3019 innermost label enclosing the `break` expression, but must enclose it.
3021 A `continue` expression is only permitted in the body of a loop.
3025 A `while` loop begins by evaluating the boolean loop conditional expression.
3026 If the loop conditional expression evaluates to `true`, the loop body block
3027 executes and control returns to the loop conditional expression. If the loop
3028 conditional expression evaluates to `false`, the `while` expression completes.
3041 Like `loop` expressions, `while` loops can be controlled with `break` or
3042 `continue`, and may optionally have a _label_. See [infinite
3043 loops](#infinite-loops), [break expressions](#break-expressions), and
3044 [continue expressions](#continue-expressions) for more information.
3046 ### `for` expressions
3048 A `for` expression is a syntactic construct for looping over elements provided
3049 by an implementation of `std::iter::IntoIterator`.
3051 An example of a `for` loop over the contents of an array:
3055 # fn bar(f: &Foo) { }
3060 let v: &[Foo] = &[a, b, c];
3067 An example of a for loop over a series of integers:
3070 # fn bar(b:usize) { }
3076 Like `loop` expressions, `for` loops can be controlled with `break` or
3077 `continue`, and may optionally have a _label_. See [infinite
3078 loops](#infinite-loops), [break expressions](#break-expressions), and
3079 [continue expressions](#continue-expressions) for more information.
3081 ### `if` expressions
3083 An `if` expression is a conditional branch in program control. The form of an
3084 `if` expression is a condition expression, followed by a consequent block, any
3085 number of `else if` conditions and blocks, and an optional trailing `else`
3086 block. The condition expressions must have type `bool`. If a condition
3087 expression evaluates to `true`, the consequent block is executed and any
3088 subsequent `else if` or `else` block is skipped. If a condition expression
3089 evaluates to `false`, the consequent block is skipped and any subsequent `else
3090 if` condition is evaluated. If all `if` and `else if` conditions evaluate to
3091 `false` then any `else` block is executed.
3093 ### `match` expressions
3095 A `match` expression branches on a *pattern*. The exact form of matching that
3096 occurs depends on the pattern. Patterns consist of some combination of
3097 literals, destructured arrays or enum constructors, structs and tuples,
3098 variable binding specifications, wildcards (`..`), and placeholders (`_`). A
3099 `match` expression has a *head expression*, which is the value to compare to
3100 the patterns. The type of the patterns must equal the type of the head
3103 In a pattern whose head expression has an `enum` type, a placeholder (`_`)
3104 stands for a *single* data field, whereas a wildcard `..` stands for *all* the
3105 fields of a particular variant.
3107 A `match` behaves differently depending on whether or not the head expression
3108 is an [lvalue or an rvalue](#lvalues-rvalues-and-temporaries). If the head
3109 expression is an rvalue, it is first evaluated into a temporary location, and
3110 the resulting value is sequentially compared to the patterns in the arms until
3111 a match is found. The first arm with a matching pattern is chosen as the branch
3112 target of the `match`, any variables bound by the pattern are assigned to local
3113 variables in the arm's block, and control enters the block.
3115 When the head expression is an lvalue, the match does not allocate a temporary
3116 location (however, a by-value binding may copy or move from the lvalue). When
3117 possible, it is preferable to match on lvalues, as the lifetime of these
3118 matches inherits the lifetime of the lvalue, rather than being restricted to
3119 the inside of the match.
3121 An example of a `match` expression:
3127 1 => println!("one"),
3128 2 => println!("two"),
3129 3 => println!("three"),
3130 4 => println!("four"),
3131 5 => println!("five"),
3132 _ => println!("something else"),
3136 Patterns that bind variables default to binding to a copy or move of the
3137 matched value (depending on the matched value's type). This can be changed to
3138 bind to a reference by using the `ref` keyword, or to a mutable reference using
3141 Subpatterns can also be bound to variables by the use of the syntax `variable @
3142 subpattern`. For example:
3148 e @ 1 ... 5 => println!("got a range element {}", e),
3149 _ => println!("anything"),
3153 Patterns can also dereference pointers by using the `&`, `&mut` and `box`
3154 symbols, as appropriate. For example, these two matches on `x: &i32` are
3159 let y = match *x { 0 => "zero", _ => "some" };
3160 let z = match x { &0 => "zero", _ => "some" };
3165 Multiple match patterns may be joined with the `|` operator. A range of values
3166 may be specified with `...`. For example:
3171 let message = match x {
3172 0 | 1 => "not many",
3178 Range patterns only work on scalar types (like integers and characters; not
3179 like arrays and structs, which have sub-components). A range pattern may not
3180 be a sub-range of another range pattern inside the same `match`.
3182 Finally, match patterns can accept *pattern guards* to further refine the
3183 criteria for matching a case. Pattern guards appear after the pattern and
3184 consist of a bool-typed expression following the `if` keyword. A pattern guard
3185 may refer to the variables bound within the pattern they follow.
3188 # let maybe_digit = Some(0);
3189 # fn process_digit(i: i32) { }
3190 # fn process_other(i: i32) { }
3192 let message = match maybe_digit {
3193 Some(x) if x < 10 => process_digit(x),
3194 Some(x) => process_other(x),
3199 ### `if let` expressions
3201 An `if let` expression is semantically identical to an `if` expression but in place
3202 of a condition expression it expects a refutable let statement. If the value of the
3203 expression on the right hand side of the let statement matches the pattern, the corresponding
3204 block will execute, otherwise flow proceeds to the first `else` block that follows.
3207 let dish = ("Ham", "Eggs");
3209 // this body will be skipped because the pattern is refuted
3210 if let ("Bacon", b) = dish {
3211 println!("Bacon is served with {}", b);
3214 // this body will execute
3215 if let ("Ham", b) = dish {
3216 println!("Ham is served with {}", b);
3220 ### `while let` loops
3222 A `while let` loop is semantically identical to a `while` loop but in place of a
3223 condition expression it expects a refutable let statement. If the value of the
3224 expression on the right hand side of the let statement matches the pattern, the
3225 loop body block executes and control returns to the pattern matching statement.
3226 Otherwise, the while expression completes.
3228 ### `return` expressions
3230 Return expressions are denoted with the keyword `return`. Evaluating a `return`
3231 expression moves its argument into the designated output location for the
3232 current function call, destroys the current function activation frame, and
3233 transfers control to the caller frame.
3235 An example of a `return` expression:
3238 fn max(a: i32, b: i32) -> i32 {
3250 Every variable, item and value in a Rust program has a type. The _type_ of a
3251 *value* defines the interpretation of the memory holding it.
3253 Built-in types and type-constructors are tightly integrated into the language,
3254 in nontrivial ways that are not possible to emulate in user-defined types.
3255 User-defined types have limited capabilities.
3259 The primitive types are the following:
3261 * The boolean type `bool` with values `true` and `false`.
3262 * The machine types (integer and floating-point).
3263 * The machine-dependent integer types.
3267 The machine types are the following:
3269 * The unsigned word types `u8`, `u16`, `u32` and `u64`, with values drawn from
3270 the integer intervals [0, 2^8 - 1], [0, 2^16 - 1], [0, 2^32 - 1] and
3271 [0, 2^64 - 1] respectively.
3273 * The signed two's complement word types `i8`, `i16`, `i32` and `i64`, with
3274 values drawn from the integer intervals [-(2^(7)), 2^7 - 1],
3275 [-(2^(15)), 2^15 - 1], [-(2^(31)), 2^31 - 1], [-(2^(63)), 2^63 - 1]
3278 * The IEEE 754-2008 `binary32` and `binary64` floating-point types: `f32` and
3279 `f64`, respectively.
3281 #### Machine-dependent integer types
3283 The `usize` type is an unsigned integer type with the same number of bits as the
3284 platform's pointer type. It can represent every memory address in the process.
3286 The `isize` type is a signed integer type with the same number of bits as the
3287 platform's pointer type. The theoretical upper bound on object and array size
3288 is the maximum `isize` value. This ensures that `isize` can be used to calculate
3289 differences between pointers into an object or array and can address every byte
3290 within an object along with one byte past the end.
3294 The types `char` and `str` hold textual data.
3296 A value of type `char` is a [Unicode scalar value](
3297 http://www.unicode.org/glossary/#unicode_scalar_value) (i.e. a code point that
3298 is not a surrogate), represented as a 32-bit unsigned word in the 0x0000 to
3299 0xD7FF or 0xE000 to 0x10FFFF range. A `[char]` array is effectively an UCS-4 /
3302 A value of type `str` is a Unicode string, represented as an array of 8-bit
3303 unsigned bytes holding a sequence of UTF-8 code points. Since `str` is of
3304 unknown size, it is not a _first-class_ type, but can only be instantiated
3305 through a pointer type, such as `&str`.
3309 A tuple *type* is a heterogeneous product of other types, called the *elements*
3310 of the tuple. It has no nominal name and is instead structurally typed.
3312 Tuple types and values are denoted by listing the types or values of their
3313 elements, respectively, in a parenthesized, comma-separated list.
3315 Because tuple elements don't have a name, they can only be accessed by
3316 pattern-matching or by using `N` directly as a field to access the
3319 An example of a tuple type and its use:
3322 type Pair<'a> = (i32, &'a str);
3323 let p: Pair<'static> = (10, "ten");
3327 assert_eq!(b, "ten");
3328 assert_eq!(p.0, 10);
3329 assert_eq!(p.1, "ten");
3332 For historical reasons and convenience, the tuple type with no elements (`()`)
3333 is often called ‘unit’ or ‘the unit type’.
3335 ### Array, and Slice types
3337 Rust has two different types for a list of items:
3339 * `[T; N]`, an 'array'
3342 An array has a fixed size, and can be allocated on either the stack or the
3345 A slice is a 'view' into an array. It doesn't own the data it points
3351 // A stack-allocated array
3352 let array: [i32; 3] = [1, 2, 3];
3354 // A heap-allocated array
3355 let vector: Vec<i32> = vec![1, 2, 3];
3357 // A slice into an array
3358 let slice: &[i32] = &vector[..];
3361 As you can see, the `vec!` macro allows you to create a `Vec<T>` easily. The
3362 `vec!` macro is also part of the standard library, rather than the language.
3364 All in-bounds elements of arrays and slices are always initialized, and access
3365 to an array or slice is always bounds-checked.
3369 A `struct` *type* is a heterogeneous product of other types, called the
3370 *fields* of the type.[^structtype]
3372 [^structtype]: `struct` types are analogous to `struct` types in C,
3373 the *record* types of the ML family,
3374 or the *struct* types of the Lisp family.
3376 New instances of a `struct` can be constructed with a [struct
3377 expression](#struct-expressions).
3379 The memory layout of a `struct` is undefined by default to allow for compiler
3380 optimizations like field reordering, but it can be fixed with the
3381 `#[repr(...)]` attribute. In either case, fields may be given in any order in
3382 a corresponding struct *expression*; the resulting `struct` value will always
3383 have the same memory layout.
3385 The fields of a `struct` may be qualified by [visibility
3386 modifiers](#visibility-and-privacy), to allow access to data in a
3387 struct outside a module.
3389 A _tuple struct_ type is just like a struct type, except that the fields are
3392 A _unit-like struct_ type is like a struct type, except that it has no
3393 fields. The one value constructed by the associated [struct
3394 expression](#struct-expressions) is the only value that inhabits such a
3397 ### Enumerated types
3399 An *enumerated type* is a nominal, heterogeneous disjoint union type, denoted
3400 by the name of an [`enum` item](#enumerations). [^enumtype]
3402 [^enumtype]: The `enum` type is analogous to a `data` constructor declaration in
3403 ML, or a *pick ADT* in Limbo.
3405 An [`enum` item](#enumerations) declares both the type and a number of *variant
3406 constructors*, each of which is independently named and takes an optional tuple
3409 New instances of an `enum` can be constructed by calling one of the variant
3410 constructors, in a [call expression](#call-expressions).
3412 Any `enum` value consumes as much memory as the largest variant constructor for
3413 its corresponding `enum` type.
3415 Enum types cannot be denoted *structurally* as types, but must be denoted by
3416 named reference to an [`enum` item](#enumerations).
3420 Nominal types — [enumerations](#enumerated-types) and
3421 [structs](#struct-types) — may be recursive. That is, each `enum`
3422 constructor or `struct` field may refer, directly or indirectly, to the
3423 enclosing `enum` or `struct` type itself. Such recursion has restrictions:
3425 * Recursive types must include a nominal type in the recursion
3426 (not mere [type definitions](grammar.html#type-definitions),
3427 or other structural types such as [arrays](#array-and-slice-types) or [tuples](#tuple-types)).
3428 * A recursive `enum` item must have at least one non-recursive constructor
3429 (in order to give the recursion a basis case).
3430 * The size of a recursive type must be finite;
3431 in other words the recursive fields of the type must be [pointer types](#pointer-types).
3432 * Recursive type definitions can cross module boundaries, but not module *visibility* boundaries,
3433 or crate boundaries (in order to simplify the module system and type checker).
3435 An example of a *recursive* type and its use:
3440 Cons(T, Box<List<T>>)
3443 let a: List<i32> = List::Cons(7, Box::new(List::Cons(13, Box::new(List::Nil))));
3448 All pointers in Rust are explicit first-class values. They can be copied,
3449 stored into data structs, and returned from functions. There are two
3450 varieties of pointer in Rust:
3453 : These point to memory _owned by some other value_.
3454 A reference type is written `&type`,
3455 or `&'a type` when you need to specify an explicit lifetime.
3456 Copying a reference is a "shallow" operation:
3457 it involves only copying the pointer itself.
3458 Releasing a reference has no effect on the value it points to,
3459 but a reference of a temporary value will keep it alive during the scope
3460 of the reference itself.
3462 * Raw pointers (`*`)
3463 : Raw pointers are pointers without safety or liveness guarantees.
3464 Raw pointers are written as `*const T` or `*mut T`,
3465 for example `*const i32` means a raw pointer to a 32-bit integer.
3466 Copying or dropping a raw pointer has no effect on the lifecycle of any
3467 other value. Dereferencing a raw pointer or converting it to any other
3468 pointer type is an [`unsafe` operation](#unsafe-functions).
3469 Raw pointers are generally discouraged in Rust code;
3470 they exist to support interoperability with foreign code,
3471 and writing performance-critical or low-level functions.
3473 The standard library contains additional 'smart pointer' types beyond references
3478 The function type constructor `fn` forms new function types. A function type
3479 consists of a possibly-empty set of function-type modifiers (such as `unsafe`
3480 or `extern`), a sequence of input types and an output type.
3482 An example of a `fn` type:
3485 fn add(x: i32, y: i32) -> i32 {
3489 let mut x = add(5,7);
3491 type Binop = fn(i32, i32) -> i32;
3492 let bo: Binop = add;
3496 #### Function types for specific items
3498 Internal to the compiler, there are also function types that are specific to a particular
3499 function item. In the following snippet, for example, the internal types of the functions
3500 `foo` and `bar` are different, despite the fact that they have the same signature:
3507 The types of `foo` and `bar` can both be implicitly coerced to the fn
3508 pointer type `fn()`. There is currently no syntax for unique fn types,
3509 though the compiler will emit a type like `fn() {foo}` in error
3510 messages to indicate "the unique fn type for the function `foo`".
3514 A [lambda expression](#lambda-expressions) produces a closure value with
3515 a unique, anonymous type that cannot be written out.
3517 Depending on the requirements of the closure, its type implements one or
3518 more of the closure traits:
3521 : The closure can be called once. A closure called as `FnOnce`
3522 can move out values from its environment.
3525 : The closure can be called multiple times as mutable. A closure called as
3526 `FnMut` can mutate values from its environment. `FnMut` inherits from
3527 `FnOnce` (i.e. anything implementing `FnMut` also implements `FnOnce`).
3530 : The closure can be called multiple times through a shared reference.
3531 A closure called as `Fn` can neither move out from nor mutate values
3532 from its environment. `Fn` inherits from `FnMut`, which itself
3533 inherits from `FnOnce`.
3538 In Rust, a type like `&SomeTrait` or `Box<SomeTrait>` is called a _trait object_.
3539 Each instance of a trait object includes:
3541 - a pointer to an instance of a type `T` that implements `SomeTrait`
3542 - a _virtual method table_, often just called a _vtable_, which contains, for
3543 each method of `SomeTrait` that `T` implements, a pointer to `T`'s
3544 implementation (i.e. a function pointer).
3546 The purpose of trait objects is to permit "late binding" of methods. A call to
3547 a method on a trait object is only resolved to a vtable entry at compile time.
3548 The actual implementation for each vtable entry can vary on an object-by-object
3551 Note that for a trait object to be instantiated, the trait must be
3552 _object-safe_. Object safety rules are defined in [RFC 255].
3554 [RFC 255]: https://github.com/rust-lang/rfcs/blob/master/text/0255-object-safety.md
3556 Given a pointer-typed expression `E` of type `&T` or `Box<T>`, where `T`
3557 implements trait `R`, casting `E` to the corresponding pointer type `&R` or
3558 `Box<R>` results in a value of the _trait object_ `R`. This result is
3559 represented as a pair of pointers: the vtable pointer for the `T`
3560 implementation of `R`, and the pointer value of `E`.
3562 An example of a trait object:
3566 fn stringify(&self) -> String;
3569 impl Printable for i32 {
3570 fn stringify(&self) -> String { self.to_string() }
3573 fn print(a: Box<Printable>) {
3574 println!("{}", a.stringify());
3578 print(Box::new(10) as Box<Printable>);
3582 In this example, the trait `Printable` occurs as a trait object in both the
3583 type signature of `print`, and the cast expression in `main`.
3587 Within the body of an item that has type parameter declarations, the names of
3588 its type parameters are types:
3591 fn to_vec<A: Clone>(xs: &[A]) -> Vec<A> {
3595 let first: A = xs[0].clone();
3596 let mut rest: Vec<A> = to_vec(&xs[1..]);
3597 rest.insert(0, first);
3602 Here, `first` has type `A`, referring to `to_vec`'s `A` type parameter; and `rest`
3603 has type `Vec<A>`, a vector with element type `A`.
3607 The special type `Self` has a meaning within traits and impls. In a trait definition, it refers
3608 to an implicit type parameter representing the "implementing" type. In an impl,
3609 it is an alias for the implementing type. For example, in:
3613 fn make_string(&self) -> String;
3616 impl Printable for String {
3617 fn make_string(&self) -> String {
3623 The notation `&self` is a shorthand for `self: &Self`. In this case,
3624 in the impl, `Self` refers to the value of type `String` that is the
3625 receiver for a call to the method `make_string`.
3629 Subtyping is implicit and can occur at any stage in type checking or
3630 inference. Subtyping in Rust is very restricted and occurs only due to
3631 variance with respect to lifetimes and between types with higher ranked
3632 lifetimes. If we were to erase lifetimes from types, then the only subtyping
3633 would be due to type equality.
3635 Consider the following example: string literals always have `'static`
3636 lifetime. Nevertheless, we can assign `s` to `t`:
3640 let s: &'static str = "hi";
3644 Since `'static` "lives longer" than `'a`, `&'static str` is a subtype of
3649 Coercions are defined in [RFC401]. A coercion is implicit and has no syntax.
3651 [RFC401]: https://github.com/rust-lang/rfcs/blob/master/text/0401-coercions.md
3655 A coercion can only occur at certain coercion sites in a program; these are
3656 typically places where the desired type is explicit or can be derived by
3657 propagation from explicit types (without type inference). Possible coercion
3660 * `let` statements where an explicit type is given.
3662 For example, `128` is coerced to have type `i8` in the following:
3668 * `static` and `const` statements (similar to `let` statements).
3670 * Arguments for function calls
3672 The value being coerced is the actual parameter, and it is coerced to
3673 the type of the formal parameter.
3675 For example, `128` is coerced to have type `i8` in the following:
3685 * Instantiations of struct or variant fields
3687 For example, `128` is coerced to have type `i8` in the following:
3690 struct Foo { x: i8 }
3697 * Function results, either the final line of a block if it is not
3698 semicolon-terminated or any expression in a `return` statement
3700 For example, `128` is coerced to have type `i8` in the following:
3708 If the expression in one of these coercion sites is a coercion-propagating
3709 expression, then the relevant sub-expressions in that expression are also
3710 coercion sites. Propagation recurses from these new coercion sites.
3711 Propagating expressions and their relevant sub-expressions are:
3713 * Array literals, where the array has type `[U; n]`. Each sub-expression in
3714 the array literal is a coercion site for coercion to type `U`.
3716 * Array literals with repeating syntax, where the array has type `[U; n]`. The
3717 repeated sub-expression is a coercion site for coercion to type `U`.
3719 * Tuples, where a tuple is a coercion site to type `(U_0, U_1, ..., U_n)`.
3720 Each sub-expression is a coercion site to the respective type, e.g. the
3721 zeroth sub-expression is a coercion site to type `U_0`.
3723 * Parenthesized sub-expressions (`(e)`): if the expression has type `U`, then
3724 the sub-expression is a coercion site to `U`.
3726 * Blocks: if a block has type `U`, then the last expression in the block (if
3727 it is not semicolon-terminated) is a coercion site to `U`. This includes
3728 blocks which are part of control flow statements, such as `if`/`else`, if
3729 the block has a known type.
3733 Coercion is allowed between the following types:
3735 * `T` to `U` if `T` is a subtype of `U` (*reflexive case*)
3737 * `T_1` to `T_3` where `T_1` coerces to `T_2` and `T_2` coerces to `T_3`
3740 Note that this is not fully supported yet
3744 * `*mut T` to `*const T`
3746 * `&T` to `*const T`
3748 * `&mut T` to `*mut T`
3750 * `&T` to `&U` if `T` implements `Deref<Target = U>`. For example:
3753 use std::ops::Deref;
3755 struct CharContainer {
3759 impl Deref for CharContainer {
3762 fn deref<'a>(&'a self) -> &'a char {
3767 fn foo(arg: &char) {}
3770 let x = &mut CharContainer { value: 'y' };
3771 foo(x); //&mut CharContainer is coerced to &char.
3775 * `&mut T` to `&mut U` if `T` implements `DerefMut<Target = U>`.
3777 * TyCtor(`T`) to TyCtor(coerce_inner(`T`)), where TyCtor(`T`) is one of
3785 - coerce_inner(`[T, ..n]`) = `[T]`
3786 - coerce_inner(`T`) = `U` where `T` is a concrete type which implements the
3789 In the future, coerce_inner will be recursively extended to tuples and
3790 structs. In addition, coercions from sub-traits to super-traits will be
3791 added. See [RFC401] for more details.
3795 Several traits define special evaluation behavior.
3799 The `Copy` trait changes the semantics of a type implementing it. Values whose
3800 type implements `Copy` are copied rather than moved upon assignment.
3802 ## The `Sized` trait
3804 The `Sized` trait indicates that the size of this type is known at compile-time.
3808 The `Drop` trait provides a destructor, to be run whenever a value of this type
3811 ## The `Deref` trait
3813 The `Deref<Target = U>` trait allows a type to implicitly implement all the methods
3814 of the type `U`. When attempting to resolve a method call, the compiler will search
3815 the top-level type for the implementation of the called method. If no such method is
3816 found, `.deref()` is called and the compiler continues to search for the method
3817 implementation in the returned type `U`.
3821 A Rust program's memory consists of a static set of *items* and a *heap*.
3822 Immutable portions of the heap may be safely shared between threads, mutable
3823 portions may not be safely shared, but several mechanisms for effectively-safe
3824 sharing of mutable values, built on unsafe code but enforcing a safe locking
3825 discipline, exist in the standard library.
3827 Allocations in the stack consist of *variables*, and allocations in the heap
3830 ### Memory allocation and lifetime
3832 The _items_ of a program are those functions, modules and types that have their
3833 value calculated at compile-time and stored uniquely in the memory image of the
3834 rust process. Items are neither dynamically allocated nor freed.
3836 The _heap_ is a general term that describes boxes. The lifetime of an
3837 allocation in the heap depends on the lifetime of the box values pointing to
3838 it. Since box values may themselves be passed in and out of frames, or stored
3839 in the heap, heap allocations may outlive the frame they are allocated within.
3841 ### Memory ownership
3843 When a stack frame is exited, its local allocations are all released, and its
3844 references to boxes are dropped.
3848 A _variable_ is a component of a stack frame, either a named function parameter,
3849 an anonymous [temporary](#lvalues-rvalues-and-temporaries), or a named local
3852 A _local variable_ (or *stack-local* allocation) holds a value directly,
3853 allocated within the stack's memory. The value is a part of the stack frame.
3855 Local variables are immutable unless declared otherwise like: `let mut x = ...`.
3857 Function parameters are immutable unless declared with `mut`. The `mut` keyword
3858 applies only to the following parameter (so `|mut x, y|` and `fn f(mut x:
3859 Box<i32>, y: Box<i32>)` declare one mutable variable `x` and one immutable
3862 Methods that take either `self` or `Box<Self>` can optionally place them in a
3863 mutable variable by prefixing them with `mut` (similar to regular arguments):
3867 fn change(mut self) -> Self;
3868 fn modify(mut self: Box<Self>) -> Box<Self>;
3872 Local variables are not initialized when allocated; the entire frame worth of
3873 local variables are allocated at once, on frame-entry, in an uninitialized
3874 state. Subsequent statements within a function may or may not initialize the
3875 local variables. Local variables can be used only after they have been
3876 initialized; this is enforced by the compiler.
3880 The Rust compiler supports various methods to link crates together both
3881 statically and dynamically. This section will explore the various methods to
3882 link Rust crates together, and more information about native libraries can be
3883 found in the [FFI section of the book][ffi].
3885 In one session of compilation, the compiler can generate multiple artifacts
3886 through the usage of either command line flags or the `crate_type` attribute.
3887 If one or more command line flags are specified, all `crate_type` attributes will
3888 be ignored in favor of only building the artifacts specified by command line.
3890 * `--crate-type=bin`, `#[crate_type = "bin"]` - A runnable executable will be
3891 produced. This requires that there is a `main` function in the crate which
3892 will be run when the program begins executing. This will link in all Rust and
3893 native dependencies, producing a distributable binary.
3895 * `--crate-type=lib`, `#[crate_type = "lib"]` - A Rust library will be produced.
3896 This is an ambiguous concept as to what exactly is produced because a library
3897 can manifest itself in several forms. The purpose of this generic `lib` option
3898 is to generate the "compiler recommended" style of library. The output library
3899 will always be usable by rustc, but the actual type of library may change from
3900 time-to-time. The remaining output types are all different flavors of
3901 libraries, and the `lib` type can be seen as an alias for one of them (but the
3902 actual one is compiler-defined).
3904 * `--crate-type=dylib`, `#[crate_type = "dylib"]` - A dynamic Rust library will
3905 be produced. This is different from the `lib` output type in that this forces
3906 dynamic library generation. The resulting dynamic library can be used as a
3907 dependency for other libraries and/or executables. This output type will
3908 create `*.so` files on linux, `*.dylib` files on osx, and `*.dll` files on
3911 * `--crate-type=staticlib`, `#[crate_type = "staticlib"]` - A static system
3912 library will be produced. This is different from other library outputs in that
3913 the Rust compiler will never attempt to link to `staticlib` outputs. The
3914 purpose of this output type is to create a static library containing all of
3915 the local crate's code along with all upstream dependencies. The static
3916 library is actually a `*.a` archive on linux and osx and a `*.lib` file on
3917 windows. This format is recommended for use in situations such as linking
3918 Rust code into an existing non-Rust application because it will not have
3919 dynamic dependencies on other Rust code.
3921 * `--crate-type=rlib`, `#[crate_type = "rlib"]` - A "Rust library" file will be
3922 produced. This is used as an intermediate artifact and can be thought of as a
3923 "static Rust library". These `rlib` files, unlike `staticlib` files, are
3924 interpreted by the Rust compiler in future linkage. This essentially means
3925 that `rustc` will look for metadata in `rlib` files like it looks for metadata
3926 in dynamic libraries. This form of output is used to produce statically linked
3927 executables as well as `staticlib` outputs.
3929 Note that these outputs are stackable in the sense that if multiple are
3930 specified, then the compiler will produce each form of output at once without
3931 having to recompile. However, this only applies for outputs specified by the
3932 same method. If only `crate_type` attributes are specified, then they will all
3933 be built, but if one or more `--crate-type` command line flags are specified,
3934 then only those outputs will be built.
3936 With all these different kinds of outputs, if crate A depends on crate B, then
3937 the compiler could find B in various different forms throughout the system. The
3938 only forms looked for by the compiler, however, are the `rlib` format and the
3939 dynamic library format. With these two options for a dependent library, the
3940 compiler must at some point make a choice between these two formats. With this
3941 in mind, the compiler follows these rules when determining what format of
3942 dependencies will be used:
3944 1. If a static library is being produced, all upstream dependencies are
3945 required to be available in `rlib` formats. This requirement stems from the
3946 reason that a dynamic library cannot be converted into a static format.
3948 Note that it is impossible to link in native dynamic dependencies to a static
3949 library, and in this case warnings will be printed about all unlinked native
3950 dynamic dependencies.
3952 2. If an `rlib` file is being produced, then there are no restrictions on what
3953 format the upstream dependencies are available in. It is simply required that
3954 all upstream dependencies be available for reading metadata from.
3956 The reason for this is that `rlib` files do not contain any of their upstream
3957 dependencies. It wouldn't be very efficient for all `rlib` files to contain a
3958 copy of `libstd.rlib`!
3960 3. If an executable is being produced and the `-C prefer-dynamic` flag is not
3961 specified, then dependencies are first attempted to be found in the `rlib`
3962 format. If some dependencies are not available in an rlib format, then
3963 dynamic linking is attempted (see below).
3965 4. If a dynamic library or an executable that is being dynamically linked is
3966 being produced, then the compiler will attempt to reconcile the available
3967 dependencies in either the rlib or dylib format to create a final product.
3969 A major goal of the compiler is to ensure that a library never appears more
3970 than once in any artifact. For example, if dynamic libraries B and C were
3971 each statically linked to library A, then a crate could not link to B and C
3972 together because there would be two copies of A. The compiler allows mixing
3973 the rlib and dylib formats, but this restriction must be satisfied.
3975 The compiler currently implements no method of hinting what format a library
3976 should be linked with. When dynamically linking, the compiler will attempt to
3977 maximize dynamic dependencies while still allowing some dependencies to be
3978 linked in via an rlib.
3980 For most situations, having all libraries available as a dylib is recommended
3981 if dynamically linking. For other situations, the compiler will emit a
3982 warning if it is unable to determine which formats to link each library with.
3984 In general, `--crate-type=bin` or `--crate-type=lib` should be sufficient for
3985 all compilation needs, and the other options are just available if more
3986 fine-grained control is desired over the output format of a Rust crate.
3990 Unsafe operations are those that potentially violate the memory-safety
3991 guarantees of Rust's static semantics.
3993 The following language level features cannot be used in the safe subset of
3996 - Dereferencing a [raw pointer](#pointer-types).
3997 - Reading or writing a [mutable static variable](#mutable-statics).
3998 - Calling an unsafe function (including an intrinsic or foreign function).
4002 Unsafe functions are functions that are not safe in all contexts and/or for all
4003 possible inputs. Such a function must be prefixed with the keyword `unsafe` and
4004 can only be called from an `unsafe` block or another `unsafe` function.
4008 A block of code can be prefixed with the `unsafe` keyword, to permit calling
4009 `unsafe` functions or dereferencing raw pointers within a safe function.
4011 When a programmer has sufficient conviction that a sequence of potentially
4012 unsafe operations is actually safe, they can encapsulate that sequence (taken
4013 as a whole) within an `unsafe` block. The compiler will consider uses of such
4014 code safe, in the surrounding context.
4016 Unsafe blocks are used to wrap foreign libraries, make direct use of hardware
4017 or implement features not directly present in the language. For example, Rust
4018 provides the language features necessary to implement memory-safe concurrency
4019 in the language but the implementation of threads and message passing is in the
4022 Rust's type system is a conservative approximation of the dynamic safety
4023 requirements, so in some cases there is a performance cost to using safe code.
4024 For example, a doubly-linked list is not a tree structure and can only be
4025 represented with reference-counted pointers in safe code. By using `unsafe`
4026 blocks to represent the reverse links as raw pointers, it can be implemented
4029 ## Behavior considered undefined
4031 The following is a list of behavior which is forbidden in all Rust code,
4032 including within `unsafe` blocks and `unsafe` functions. Type checking provides
4033 the guarantee that these issues are never caused by safe code.
4036 * Dereferencing a null/dangling raw pointer
4037 * Reads of [undef](http://llvm.org/docs/LangRef.html#undefined-values)
4038 (uninitialized) memory
4039 * Breaking the [pointer aliasing
4040 rules](http://llvm.org/docs/LangRef.html#pointer-aliasing-rules)
4041 with raw pointers (a subset of the rules used by C)
4042 * `&mut` and `&` follow LLVM’s scoped [noalias] model, except if the `&T`
4043 contains an `UnsafeCell<U>`. Unsafe code must not violate these aliasing
4045 * Mutating non-mutable data (that is, data reached through a shared reference or
4046 data owned by a `let` binding), unless that data is contained within an `UnsafeCell<U>`.
4047 * Invoking undefined behavior via compiler intrinsics:
4048 * Indexing outside of the bounds of an object with `std::ptr::offset`
4049 (`offset` intrinsic), with
4050 the exception of one byte past the end which is permitted.
4051 * Using `std::ptr::copy_nonoverlapping_memory` (`memcpy32`/`memcpy64`
4052 intrinsics) on overlapping buffers
4053 * Invalid values in primitive types, even in private fields/locals:
4054 * Dangling/null references or boxes
4055 * A value other than `false` (0) or `true` (1) in a `bool`
4056 * A discriminant in an `enum` not included in the type definition
4057 * A value in a `char` which is a surrogate or above `char::MAX`
4058 * Non-UTF-8 byte sequences in a `str`
4059 * Unwinding into Rust from foreign code or unwinding from Rust into foreign
4060 code. Rust's failure system is not compatible with exception handling in
4061 other languages. Unwinding must be caught and handled at FFI boundaries.
4063 [noalias]: http://llvm.org/docs/LangRef.html#noalias
4065 ## Behavior not considered unsafe
4067 This is a list of behavior not considered *unsafe* in Rust terms, but that may
4071 * Leaks of memory and other resources
4072 * Exiting without calling destructors
4074 - Overflow is considered "unexpected" behavior and is always user-error,
4075 unless the `wrapping` primitives are used. In non-optimized builds, the compiler
4076 will insert debug checks that panic on overflow, but in optimized builds overflow
4077 instead results in wrapped values. See [RFC 560] for the rationale and more details.
4079 [RFC 560]: https://github.com/rust-lang/rfcs/blob/master/text/0560-integer-overflow.md
4081 # Appendix: Influences
4083 Rust is not a particularly original language, with design elements coming from
4084 a wide range of sources. Some of these are listed below (including elements
4085 that have since been removed):
4087 * SML, OCaml: algebraic data types, pattern matching, type inference,
4088 semicolon statement separation
4089 * C++: references, RAII, smart pointers, move semantics, monomorphization,
4091 * ML Kit, Cyclone: region based memory management
4092 * Haskell (GHC): typeclasses, type families
4093 * Newsqueak, Alef, Limbo: channels, concurrency
4094 * Erlang: message passing, thread failure, ~~linked thread failure~~,
4095 ~~lightweight concurrency~~
4096 * Swift: optional bindings
4097 * Scheme: hygienic macros
4099 * Ruby: ~~block syntax~~
4100 * NIL, Hermes: ~~typestate~~
4101 * [Unicode Annex #31](http://www.unicode.org/reports/tr31/): identifier and
4104 [ffi]: book/ffi.html
4105 [plugin]: book/compiler-plugins.html