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 Finally, this document is not normative. It may include details that are
25 specific to `rustc` itself, and should not be taken as a specification for
26 the Rust language. We intend to produce such a document someday, but this
27 is what we have for now.
29 You may also be interested in the [grammar].
31 [book]: book/index.html
32 [standard]: std/index.html
33 [grammar]: grammar.html
37 ## Unicode productions
39 A few productions in Rust's grammar permit Unicode code points outside the
40 ASCII range. We define these productions in terms of character properties
41 specified in the Unicode standard, rather than in terms of ASCII-range code
42 points. The grammar has a [Special Unicode Productions][unicodeproductions]
43 section that lists these productions.
45 [unicodeproductions]: grammar.html#special-unicode-productions
47 ## String table productions
49 Some rules in the grammar — notably [unary
50 operators](#unary-operator-expressions), [binary
51 operators](#binary-operator-expressions), and [keywords][keywords] — are
52 given in a simplified form: as a listing of a table of unquoted, printable
53 whitespace-separated strings. These cases form a subset of the rules regarding
54 the [token](#tokens) rule, and are assumed to be the result of a
55 lexical-analysis phase feeding the parser, driven by a DFA, operating over the
56 disjunction of all such string table entries.
58 [keywords]: grammar.html#keywords
60 When such a string enclosed in double-quotes (`"`) occurs inside the grammar,
61 it is an implicit reference to a single member of such a string table
62 production. See [tokens](#tokens) for more information.
68 Rust input is interpreted as a sequence of Unicode code points encoded in UTF-8.
69 Most Rust grammar rules are defined in terms of printable ASCII-range
70 code points, but a small number are defined in terms of Unicode properties or
71 explicit code point lists. [^inputformat]
73 [^inputformat]: Substitute definitions for the special Unicode productions are
74 provided to the grammar verifier, restricted to ASCII range, when verifying the
75 grammar in this document.
79 An identifier is any nonempty Unicode[^non_ascii_idents] string of the following form:
81 [^non_ascii_idents]: Non-ASCII characters in identifiers are currently feature
82 gated. This is expected to improve soon.
86 * The first character has property `XID_start`
87 * The remaining characters have property `XID_continue`
91 * The first character is `_`
92 * The identifier is more than one character, `_` alone is not an identifier
93 * The remaining characters have property `XID_continue`
95 that does _not_ occur in the set of [keywords][keywords].
97 > **Note**: `XID_start` and `XID_continue` as character properties cover the
98 > character ranges used to form the more familiar C and Java language-family
103 Comments in Rust code follow the general C++ style of line (`//`) and
104 block (`/* ... */`) comment forms. Nested block comments are supported.
106 Line comments beginning with exactly _three_ slashes (`///`), and block
107 comments (`/** ... */`), are interpreted as a special syntax for `doc`
108 [attributes](#attributes). That is, they are equivalent to writing
109 `#[doc="..."]` around the body of the comment, i.e., `/// Foo` turns into
112 Line comments beginning with `//!` and block comments `/*! ... */` are
113 doc comments that apply to the parent of the comment, rather than the item
114 that follows. That is, they are equivalent to writing `#![doc="..."]` around
115 the body of the comment. `//!` comments are usually used to document
116 modules that occupy a source file.
118 Non-doc comments are interpreted as a form of whitespace.
122 Whitespace is any non-empty string containing only characters that have the
123 `Pattern_White_Space` Unicode property, namely:
125 - `U+0009` (horizontal tab, `'\t'`)
126 - `U+000A` (line feed, `'\n'`)
127 - `U+000B` (vertical tab)
128 - `U+000C` (form feed)
129 - `U+000D` (carriage return, `'\r'`)
130 - `U+0020` (space, `' '`)
131 - `U+0085` (next line)
132 - `U+200E` (left-to-right mark)
133 - `U+200F` (right-to-left mark)
134 - `U+2028` (line separator)
135 - `U+2029` (paragraph separator)
137 Rust is a "free-form" language, meaning that all forms of whitespace serve only
138 to separate _tokens_ in the grammar, and have no semantic significance.
140 A Rust program has identical meaning if each whitespace element is replaced
141 with any other legal whitespace element, such as a single space character.
145 Tokens are primitive productions in the grammar defined by regular
146 (non-recursive) languages. "Simple" tokens are given in [string table
147 production](#string-table-productions) form, and occur in the rest of the
148 grammar as double-quoted strings. Other tokens have exact rules given.
152 A literal is an expression consisting of a single token, rather than a sequence
153 of tokens, that immediately and directly denotes the value it evaluates to,
154 rather than referring to it by name or some other evaluation rule. A literal is
155 a form of constant expression, so is evaluated (primarily) at compile time.
159 ##### Characters and strings
161 | | Example | `#` sets | Characters | Escapes |
162 |----------------------------------------------|-----------------|------------|-------------|---------------------|
163 | [Character](#character-literals) | `'H'` | `N/A` | All Unicode | [Quote](#quote-escapes) & [Byte](#byte-escapes) & [Unicode](#unicode-escapes) |
164 | [String](#string-literals) | `"hello"` | `N/A` | All Unicode | [Quote](#quote-escapes) & [Byte](#byte-escapes) & [Unicode](#unicode-escapes) |
165 | [Raw](#raw-string-literals) | `r#"hello"#` | `0...` | All Unicode | `N/A` |
166 | [Byte](#byte-literals) | `b'H'` | `N/A` | All ASCII | [Quote](#quote-escapes) & [Byte](#byte-escapes) |
167 | [Byte string](#byte-string-literals) | `b"hello"` | `N/A` | All ASCII | [Quote](#quote-escapes) & [Byte](#byte-escapes) |
168 | [Raw byte string](#raw-byte-string-literals) | `br#"hello"#` | `0...` | All ASCII | `N/A` |
174 | `\x7F` | 8-bit character code (exactly 2 digits) |
176 | `\r` | Carriage return |
181 ##### Unicode escapes
184 | `\u{7FFF}` | 24-bit Unicode character code (up to 6 digits) |
189 | `\'` | Single quote |
190 | `\"` | Double quote |
194 | [Number literals](#number-literals)`*` | Example | Exponentiation | Suffixes |
195 |----------------------------------------|---------|----------------|----------|
196 | Decimal integer | `98_222` | `N/A` | Integer suffixes |
197 | Hex integer | `0xff` | `N/A` | Integer suffixes |
198 | Octal integer | `0o77` | `N/A` | Integer suffixes |
199 | Binary integer | `0b1111_0000` | `N/A` | Integer suffixes |
200 | Floating-point | `123.0E+77` | `Optional` | Floating-point suffixes |
202 `*` All number literals allow `_` as a visual separator: `1_234.0E+18f64`
205 | Integer | Floating-point |
206 |---------|----------------|
207 | `u8`, `i8`, `u16`, `i16`, `u32`, `i32`, `u64`, `i64`, `isize`, `usize` | `f32`, `f64` |
209 #### Character and string literals
211 ##### Character literals
213 A _character literal_ is a single Unicode character enclosed within two
214 `U+0027` (single-quote) characters, with the exception of `U+0027` itself,
215 which must be _escaped_ by a preceding `U+005C` character (`\`).
217 ##### String literals
219 A _string literal_ is a sequence of any Unicode characters enclosed within two
220 `U+0022` (double-quote) characters, with the exception of `U+0022` itself,
221 which must be _escaped_ by a preceding `U+005C` character (`\`).
223 Line-break characters are allowed in string literals. Normally they represent
224 themselves (i.e. no translation), but as a special exception, when an unescaped
225 `U+005C` character (`\`) occurs immediately before the newline (`U+000A`), the
226 `U+005C` character, the newline, and all whitespace at the beginning of the
227 next line are ignored. Thus `a` and `b` are equal:
237 ##### Character escapes
239 Some additional _escapes_ are available in either character or non-raw string
240 literals. An escape starts with a `U+005C` (`\`) and continues with one of the
243 * An _8-bit code point escape_ starts with `U+0078` (`x`) and is
244 followed by exactly two _hex digits_. It denotes the Unicode code point
245 equal to the provided hex value.
246 * A _24-bit code point escape_ starts with `U+0075` (`u`) and is followed
247 by up to six _hex digits_ surrounded by braces `U+007B` (`{`) and `U+007D`
248 (`}`). It denotes the Unicode code point equal to the provided hex value.
249 * A _whitespace escape_ is one of the characters `U+006E` (`n`), `U+0072`
250 (`r`), or `U+0074` (`t`), denoting the Unicode values `U+000A` (LF),
251 `U+000D` (CR) or `U+0009` (HT) respectively.
252 * The _null escape_ is the character `U+0030` (`0`) and denotes the Unicode
253 value `U+0000` (NUL).
254 * The _backslash escape_ is the character `U+005C` (`\`) which must be
255 escaped in order to denote *itself*.
257 ##### Raw string literals
259 Raw string literals do not process any escapes. They start with the character
260 `U+0072` (`r`), followed by zero or more of the character `U+0023` (`#`) and a
261 `U+0022` (double-quote) character. The _raw string body_ can contain any sequence
262 of Unicode characters and is terminated only by another `U+0022` (double-quote)
263 character, followed by the same number of `U+0023` (`#`) characters that preceded
264 the opening `U+0022` (double-quote) character.
266 All Unicode characters contained in the raw string body represent themselves,
267 the characters `U+0022` (double-quote) (except when followed by at least as
268 many `U+0023` (`#`) characters as were used to start the raw string literal) or
269 `U+005C` (`\`) do not have any special meaning.
271 Examples for string literals:
274 "foo"; r"foo"; // foo
275 "\"foo\""; r#""foo""#; // "foo"
278 r##"foo #"# bar"##; // foo #"# bar
280 "\x52"; "R"; r"R"; // R
281 "\\x52"; r"\x52"; // \x52
284 #### Byte and byte string literals
288 A _byte literal_ is a single ASCII character (in the `U+0000` to `U+007F`
289 range) or a single _escape_ preceded by the characters `U+0062` (`b`) and
290 `U+0027` (single-quote), and followed by the character `U+0027`. If the character
291 `U+0027` is present within the literal, it must be _escaped_ by a preceding
292 `U+005C` (`\`) character. It is equivalent to a `u8` unsigned 8-bit integer
295 ##### Byte string literals
297 A non-raw _byte string literal_ is a sequence of ASCII characters and _escapes_,
298 preceded by the characters `U+0062` (`b`) and `U+0022` (double-quote), and
299 followed by the character `U+0022`. If the character `U+0022` is present within
300 the literal, it must be _escaped_ by a preceding `U+005C` (`\`) character.
301 Alternatively, a byte string literal can be a _raw byte string literal_, defined
302 below. A byte string literal of length `n` is equivalent to a `&'static [u8; n]` borrowed fixed-sized array
303 of unsigned 8-bit integers.
305 Some additional _escapes_ are available in either byte or non-raw byte string
306 literals. An escape starts with a `U+005C` (`\`) and continues with one of the
309 * A _byte escape_ escape starts with `U+0078` (`x`) and is
310 followed by exactly two _hex digits_. It denotes the byte
311 equal to the provided hex value.
312 * A _whitespace escape_ is one of the characters `U+006E` (`n`), `U+0072`
313 (`r`), or `U+0074` (`t`), denoting the bytes values `0x0A` (ASCII LF),
314 `0x0D` (ASCII CR) or `0x09` (ASCII HT) respectively.
315 * The _null escape_ is the character `U+0030` (`0`) and denotes the byte
316 value `0x00` (ASCII NUL).
317 * The _backslash escape_ is the character `U+005C` (`\`) which must be
318 escaped in order to denote its ASCII encoding `0x5C`.
320 ##### Raw byte string literals
322 Raw byte string literals do not process any escapes. They start with the
323 character `U+0062` (`b`), followed by `U+0072` (`r`), followed by zero or more
324 of the character `U+0023` (`#`), and a `U+0022` (double-quote) character. The
325 _raw string body_ can contain any sequence of ASCII characters and is terminated
326 only by another `U+0022` (double-quote) character, followed by the same number of
327 `U+0023` (`#`) characters that preceded the opening `U+0022` (double-quote)
328 character. A raw byte string literal can not contain any non-ASCII byte.
330 All characters contained in the raw string body represent their ASCII encoding,
331 the characters `U+0022` (double-quote) (except when followed by at least as
332 many `U+0023` (`#`) characters as were used to start the raw string literal) or
333 `U+005C` (`\`) do not have any special meaning.
335 Examples for byte string literals:
338 b"foo"; br"foo"; // foo
339 b"\"foo\""; br#""foo""#; // "foo"
342 br##"foo #"# bar"##; // foo #"# bar
344 b"\x52"; b"R"; br"R"; // R
345 b"\\x52"; br"\x52"; // \x52
350 A _number literal_ is either an _integer literal_ or a _floating-point
351 literal_. The grammar for recognizing the two kinds of literals is mixed.
353 ##### Integer literals
355 An _integer literal_ has one of four forms:
357 * A _decimal literal_ starts with a *decimal digit* and continues with any
358 mixture of *decimal digits* and _underscores_.
359 * A _hex literal_ starts with the character sequence `U+0030` `U+0078`
360 (`0x`) and continues as any mixture of hex digits and underscores.
361 * An _octal literal_ starts with the character sequence `U+0030` `U+006F`
362 (`0o`) and continues as any mixture of octal digits and underscores.
363 * A _binary literal_ starts with the character sequence `U+0030` `U+0062`
364 (`0b`) and continues as any mixture of binary digits and underscores.
366 Like any literal, an integer literal may be followed (immediately,
367 without any spaces) by an _integer suffix_, which forcibly sets the
368 type of the literal. The integer suffix must be the name of one of the
369 integral types: `u8`, `i8`, `u16`, `i16`, `u32`, `i32`, `u64`, `i64`,
372 The type of an _unsuffixed_ integer literal is determined by type inference:
374 * If an integer type can be _uniquely_ determined from the surrounding
375 program context, the unsuffixed integer literal has that type.
377 * If the program context under-constrains the type, it defaults to the
378 signed 32-bit integer `i32`.
380 * If the program context over-constrains the type, it is considered a
383 Examples of integer literals of various forms:
390 0o70_i16; // type i16
391 0b1111_1111_1001_0000_i32; // type i32
392 0usize; // type usize
395 Note that the Rust syntax considers `-1i8` as an application of the [unary minus
396 operator](#unary-operator-expressions) to an integer literal `1i8`, rather than
397 a single integer literal.
399 ##### Floating-point literals
401 A _floating-point literal_ has one of two forms:
403 * A _decimal literal_ followed by a period character `U+002E` (`.`). This is
404 optionally followed by another decimal literal, with an optional _exponent_.
405 * A single _decimal literal_ followed by an _exponent_.
407 Like integer literals, a floating-point literal may be followed by a
408 suffix, so long as the pre-suffix part does not end with `U+002E` (`.`).
409 The suffix forcibly sets the type of the literal. There are two valid
410 _floating-point suffixes_, `f32` and `f64` (the 32-bit and 64-bit floating point
411 types), which explicitly determine the type of the literal.
413 The type of an _unsuffixed_ floating-point literal is determined by
416 * If a floating-point type can be _uniquely_ determined from the
417 surrounding program context, the unsuffixed floating-point literal
420 * If the program context under-constrains the type, it defaults to `f64`.
422 * If the program context over-constrains the type, it is considered a
425 Examples of floating-point literals of various forms:
428 123.0f64; // type f64
431 12E+99_f64; // type f64
432 let x: f64 = 2.; // type f64
435 This last example is different because it is not possible to use the suffix
436 syntax with a floating point literal ending in a period. `2.f64` would attempt
437 to call a method named `f64` on `2`.
439 The representation semantics of floating-point numbers are described in
440 ["Machine Types"](#machine-types).
442 #### Boolean literals
444 The two values of the boolean type are written `true` and `false`.
448 Symbols are a general class of printable [tokens](#tokens) that play structural
449 roles in a variety of grammar productions. They are a
450 set of remaining miscellaneous printable tokens that do not
451 otherwise appear as [unary operators](#unary-operator-expressions), [binary
452 operators](#binary-operator-expressions), or [keywords][keywords].
453 They are catalogued in [the Symbols section][symbols] of the Grammar document.
455 [symbols]: grammar.html#symbols
460 A _path_ is a sequence of one or more path components _logically_ separated by
461 a namespace qualifier (`::`). If a path consists of only one component, it may
462 refer to either an [item](#items) or a [variable](#variables) in a local control
463 scope. If a path has multiple components, it refers to an item.
465 Every item has a _canonical path_ within its crate, but the path naming an item
466 is only meaningful within a given crate. There is no global namespace across
467 crates; an item's canonical path merely identifies it within the crate.
469 Two examples of simple paths consisting of only identifier components:
476 Path components are usually [identifiers](#identifiers), but they may
477 also include angle-bracket-enclosed lists of type arguments. In
478 [expression](#expressions) context, the type argument list is given
479 after a `::` namespace qualifier in order to disambiguate it from a
480 relational expression involving the less-than symbol (`<`). In type
481 expression context, the final namespace qualifier is omitted.
483 Two examples of paths with type arguments:
486 # struct HashMap<K, V>(K,V);
488 # fn id<T>(t: T) -> T { t }
489 type T = HashMap<i32,String>; // Type arguments used in a type expression
490 let x = id::<i32>(10); // Type arguments used in a call expression
494 Paths can be denoted with various leading qualifiers to change the meaning of
497 * Paths starting with `::` are considered to be global paths where the
498 components of the path start being resolved from the crate root. Each
499 identifier in the path must resolve to an item.
507 ::a::foo(); // call a's foo function
513 * Paths starting with the keyword `super` begin resolution relative to the
514 parent module. Each further identifier must resolve to an item.
522 super::a::foo(); // call a's foo function
528 * Paths starting with the keyword `self` begin resolution relative to the
529 current module. Each further identifier must resolve to an item.
539 Additionally keyword `super` may be repeated several times after the first
540 `super` or `self` to refer to ancestor modules.
549 super::super::foo(); // call a's foo function
550 self::super::super::foo(); // call a's foo function
560 A number of minor features of Rust are not central enough to have their own
561 syntax, and yet are not implementable as functions. Instead, they are given
562 names, and invoked through a consistent syntax: `some_extension!(...)`.
564 Users of `rustc` can define new macros in two ways:
566 * [Macros](book/macros.html) define new syntax in a higher-level,
568 * [Procedural Macros][procedural macros] can be used to implement custom derive.
570 And one unstable way: [compiler plugins][plugin].
574 `macro_rules` allows users to define syntax extension in a declarative way. We
575 call such extensions "macros by example" or simply "macros".
577 Currently, macros can expand to expressions, statements, items, or patterns.
579 (A `sep_token` is any token other than `*` and `+`. A `non_special_token` is
580 any token other than a delimiter or `$`.)
582 The macro expander looks up macro invocations by name, and tries each macro
583 rule in turn. It transcribes the first successful match. Matching and
584 transcription are closely related to each other, and we will describe them
589 The macro expander matches and transcribes every token that does not begin with
590 a `$` literally, including delimiters. For parsing reasons, delimiters must be
591 balanced, but they are otherwise not special.
593 In the matcher, `$` _name_ `:` _designator_ matches the nonterminal in the Rust
594 syntax named by _designator_. Valid designators are:
596 * `item`: an [item](#items)
597 * `block`: a [block](#block-expressions)
598 * `stmt`: a [statement](#statements)
599 * `pat`: a [pattern](#match-expressions)
600 * `expr`: an [expression](#expressions)
601 * `ty`: a [type](#types)
602 * `ident`: an [identifier](#identifiers)
603 * `path`: a [path](#paths)
604 * `tt`: a token tree (a single [token](#tokens) or a sequence of token trees surrounded
605 by matching `()`, `[]`, or `{}`)
606 * `meta`: the contents of an [attribute](#attributes)
608 In the transcriber, the
609 designator is already known, and so only the name of a matched nonterminal comes
610 after the dollar sign.
612 In both the matcher and transcriber, the Kleene star-like operator indicates
613 repetition. The Kleene star operator consists of `$` and parentheses, optionally
614 followed by a separator token, followed by `*` or `+`. `*` means zero or more
615 repetitions, `+` means at least one repetition. The parentheses are not matched or
616 transcribed. On the matcher side, a name is bound to _all_ of the names it
617 matches, in a structure that mimics the structure of the repetition encountered
618 on a successful match. The job of the transcriber is to sort that structure
621 The rules for transcription of these repetitions are called "Macro By Example".
622 Essentially, one "layer" of repetition is discharged at a time, and all of them
623 must be discharged by the time a name is transcribed. Therefore, `( $( $i:ident
624 ),* ) => ( $i )` is an invalid macro, but `( $( $i:ident ),* ) => ( $( $i:ident
625 ),* )` is acceptable (if trivial).
627 When Macro By Example encounters a repetition, it examines all of the `$`
628 _name_ s that occur in its body. At the "current layer", they all must repeat
629 the same number of times, so ` ( $( $i:ident ),* ; $( $j:ident ),* ) => ( $(
630 ($i,$j) ),* )` is valid if given the argument `(a,b,c ; d,e,f)`, but not
631 `(a,b,c ; d,e)`. The repetition walks through the choices at that layer in
632 lockstep, so the former input transcribes to `(a,d), (b,e), (c,f)`.
634 Nested repetitions are allowed.
636 ### Parsing limitations
638 The parser used by the macro system is reasonably powerful, but the parsing of
639 Rust syntax is restricted in two ways:
641 1. Macro definitions are required to include suitable separators after parsing
642 expressions and other bits of the Rust grammar. This implies that
643 a macro definition like `$i:expr [ , ]` is not legal, because `[` could be part
644 of an expression. A macro definition like `$i:expr,` or `$i:expr;` would be legal,
645 however, because `,` and `;` are legal separators. See [RFC 550] for more information.
646 2. The parser must have eliminated all ambiguity by the time it reaches a `$`
647 _name_ `:` _designator_. This requirement most often affects name-designator
648 pairs when they occur at the beginning of, or immediately after, a `$(...)*`;
649 requiring a distinctive token in front can solve the problem.
651 [RFC 550]: https://github.com/rust-lang/rfcs/blob/master/text/0550-macro-future-proofing.md
655 "Procedrual macros" are the second way to implement a macro. For now, the only
656 thing they can be used for is to implement derive on your own types. See
657 [the book][procedural macros] for a tutorial.
659 Procedural macros involve a few different parts of the language and its
660 standard libraries. First is the `proc_macro` crate, included with Rust,
661 that defines an interface for building a procedrual macro. The
662 `#[proc_macro_derive(Foo)]` attribute is used to mark the the deriving
663 function. This function must have the type signature:
666 use proc_macro::TokenStream;
668 #[proc_macro_derive(Hello)]
669 pub fn hello_world(input: TokenStream) -> TokenStream
672 Finally, procedural macros must be in their own crate, with the `proc-macro`
675 # Crates and source files
677 Although Rust, like any other language, can be implemented by an interpreter as
678 well as a compiler, the only existing implementation is a compiler,
680 always been designed to be compiled. For these reasons, this section assumes a
683 Rust's semantics obey a *phase distinction* between compile-time and
684 run-time.[^phase-distinction] Semantic rules that have a *static
685 interpretation* govern the success or failure of compilation, while
687 that have a *dynamic interpretation* govern the behavior of the program at
690 [^phase-distinction]: This distinction would also exist in an interpreter.
691 Static checks like syntactic analysis, type checking, and lints should
692 happen before the program is executed regardless of when it is executed.
694 The compilation model centers on artifacts called _crates_. Each compilation
695 processes a single crate in source form, and if successful, produces a single
696 crate in binary form: either an executable or some sort of
697 library.[^cratesourcefile]
699 [^cratesourcefile]: A crate is somewhat analogous to an *assembly* in the
700 ECMA-335 CLI model, a *library* in the SML/NJ Compilation Manager, a *unit*
701 in the Owens and Flatt module system, or a *configuration* in Mesa.
703 A _crate_ is a unit of compilation and linking, as well as versioning,
704 distribution and runtime loading. A crate contains a _tree_ of nested
705 [module](#modules) scopes. The top level of this tree is a module that is
706 anonymous (from the point of view of paths within the module) and any item
707 within a crate has a canonical [module path](#paths) denoting its location
708 within the crate's module tree.
710 The Rust compiler is always invoked with a single source file as input, and
711 always produces a single output crate. The processing of that source file may
712 result in other source files being loaded as modules. Source files have the
715 A Rust source file describes a module, the name and location of which —
716 in the module tree of the current crate — are defined from outside the
717 source file: either by an explicit `mod_item` in a referencing source file, or
718 by the name of the crate itself. Every source file is a module, but not every
719 module needs its own source file: [module definitions](#modules) can be nested
722 Each source file contains a sequence of zero or more `item` definitions, and
723 may optionally begin with any number of [attributes](#items-and-attributes)
724 that apply to the containing module, most of which influence the behavior of
725 the compiler. The anonymous crate module can have additional attributes that
726 apply to the crate as a whole.
729 // Specify the crate name.
730 #![crate_name = "projx"]
732 // Specify the type of output artifact.
733 #![crate_type = "lib"]
735 // Turn on a warning.
736 // This can be done in any module, not just the anonymous crate module.
737 #![warn(non_camel_case_types)]
740 A crate that contains a `main` function can be compiled to an executable. If a
741 `main` function is present, its return type must be `()`
742 ("[unit](#tuple-types)") and it must take no arguments.
744 # Items and attributes
746 Crates contain [items](#items), each of which may have some number of
747 [attributes](#attributes) attached to it.
751 An _item_ is a component of a crate. Items are organized within a crate by a
752 nested set of [modules](#modules). Every crate has a single "outermost"
753 anonymous module; all further items within the crate have [paths](#paths)
754 within the module tree of the crate.
756 Items are entirely determined at compile-time, generally remain fixed during
757 execution, and may reside in read-only memory.
759 There are several kinds of item:
761 * [`extern crate` declarations](#extern-crate-declarations)
762 * [`use` declarations](#use-declarations)
763 * [modules](#modules)
764 * [function definitions](#functions)
765 * [`extern` blocks](#external-blocks)
766 * [type definitions](grammar.html#type-definitions)
767 * [struct definitions](#structs)
768 * [enumeration definitions](#enumerations)
769 * [constant items](#constant-items)
770 * [static items](#static-items)
771 * [trait definitions](#traits)
772 * [implementations](#implementations)
774 Some items form an implicit scope for the declaration of sub-items. In other
775 words, within a function or module, declarations of items can (in many cases)
776 be mixed with the statements, control blocks, and similar artifacts that
777 otherwise compose the item body. The meaning of these scoped items is the same
778 as if the item was declared outside the scope — it is still a static item
779 — except that the item's *path name* within the module namespace is
780 qualified by the name of the enclosing item, or is private to the enclosing
781 item (in the case of functions). The grammar specifies the exact locations in
782 which sub-item declarations may appear.
786 All items except modules, constants and statics may be *parameterized* by type.
787 Type parameters are given as a comma-separated list of identifiers enclosed in
788 angle brackets (`<...>`), after the name of the item and before its definition.
789 The type parameters of an item are considered "part of the name", not part of
790 the type of the item. A referencing [path](#paths) must (in principle) provide
791 type arguments as a list of comma-separated types enclosed within angle
792 brackets, in order to refer to the type-parameterized item. In practice, the
793 type-inference system can usually infer such argument types from context. There
794 are no general type-parametric types, only type-parametric items. That is, Rust
795 has no notion of type abstraction: there are no higher-ranked (or "forall") types
796 abstracted over other types, though higher-ranked types do exist for lifetimes.
800 A module is a container for zero or more [items](#items).
802 A _module item_ is a module, surrounded in braces, named, and prefixed with the
803 keyword `mod`. A module item introduces a new, named module into the tree of
804 modules making up a crate. Modules can nest arbitrarily.
806 An example of a module:
810 type Complex = (f64, f64);
811 fn sin(f: f64) -> f64 {
815 fn cos(f: f64) -> f64 {
819 fn tan(f: f64) -> f64 {
826 Modules and types share the same namespace. Declaring a named type with
827 the same name as a module in scope is forbidden: that is, a type definition,
828 trait, struct, enumeration, or type parameter can't shadow the name of a module
829 in scope, or vice versa.
831 A module without a body is loaded from an external file, by default with the
832 same name as the module, plus the `.rs` extension. When a nested submodule is
833 loaded from an external file, it is loaded from a subdirectory path that
834 mirrors the module hierarchy.
837 // Load the `vec` module from `vec.rs`
841 // Load the `local_data` module from `thread/local_data.rs`
842 // or `thread/local_data/mod.rs`.
847 The directories and files used for loading external file modules can be
848 influenced with the `path` attribute.
851 #[path = "thread_files"]
853 // Load the `local_data` module from `thread_files/tls.rs`
859 #### Extern crate declarations
861 An _`extern crate` declaration_ specifies a dependency on an external crate.
862 The external crate is then bound into the declaring scope as the `ident`
863 provided in the `extern_crate_decl`.
865 The external crate is resolved to a specific `soname` at compile time, and a
866 runtime linkage requirement to that `soname` is passed to the linker for
867 loading at runtime. The `soname` is resolved at compile time by scanning the
868 compiler's library path and matching the optional `crateid` provided against
869 the `crateid` attributes that were declared on the external crate when it was
870 compiled. If no `crateid` is provided, a default `name` attribute is assumed,
871 equal to the `ident` given in the `extern_crate_decl`.
873 Three examples of `extern crate` declarations:
878 extern crate std; // equivalent to: extern crate std as std;
880 extern crate std as ruststd; // linking to 'std' under another name
883 When naming Rust crates, hyphens are disallowed. However, Cargo packages may
884 make use of them. In such case, when `Cargo.toml` doesn't specify a crate name,
885 Cargo will transparently replace `-` with `_` (Refer to [RFC 940] for more
891 // Importing the Cargo package hello-world
892 extern crate hello_world; // hyphen replaced with an underscore
895 [RFC 940]: https://github.com/rust-lang/rfcs/blob/master/text/0940-hyphens-considered-harmful.md
897 #### Use declarations
899 A _use declaration_ creates one or more local name bindings synonymous with
900 some other [path](#paths). Usually a `use` declaration is used to shorten the
901 path required to refer to a module item. These declarations may appear in
902 [modules](#modules) and [blocks](grammar.html#block-expressions), usually at the top.
904 > **Note**: Unlike in many languages,
905 > `use` declarations in Rust do *not* declare linkage dependency with external crates.
906 > Rather, [`extern crate` declarations](#extern-crate-declarations) declare linkage dependencies.
908 Use declarations support a number of convenient shortcuts:
910 * Rebinding the target name as a new local name, using the syntax `use p::q::r as x;`
911 * Simultaneously binding a list of paths differing only in their final element,
912 using the glob-like brace syntax `use a::b::{c,d,e,f};`
913 * Binding all paths matching a given prefix, using the asterisk wildcard syntax
915 * Simultaneously binding a list of paths differing only in their final element
916 and their immediate parent module, using the `self` keyword, such as
917 `use a::b::{self, c, d};`
919 An example of `use` declarations:
922 use std::option::Option::{Some, None};
923 use std::collections::hash_map::{self, HashMap};
926 fn bar(map1: HashMap<String, usize>, map2: hash_map::HashMap<String, usize>){}
929 // Equivalent to 'foo(vec![std::option::Option::Some(1.0f64),
930 // std::option::Option::None]);'
931 foo(vec![Some(1.0f64), None]);
933 // Both `hash_map` and `HashMap` are in scope.
934 let map1 = HashMap::new();
935 let map2 = hash_map::HashMap::new();
940 Like items, `use` declarations are private to the containing module, by
941 default. Also like items, a `use` declaration can be public, if qualified by
942 the `pub` keyword. Such a `use` declaration serves to _re-export_ a name. A
943 public `use` declaration can therefore _redirect_ some public name to a
944 different target definition: even a definition with a private canonical path,
945 inside a different module. If a sequence of such redirections form a cycle or
946 cannot be resolved unambiguously, they represent a compile-time error.
948 An example of re-exporting:
953 pub use quux::foo::{bar, baz};
962 In this example, the module `quux` re-exports two public names defined in
965 Also note that the paths contained in `use` items are relative to the crate
966 root. So, in the previous example, the `use` refers to `quux::foo::{bar,
967 baz}`, and not simply to `foo::{bar, baz}`. This also means that top-level
968 module declarations should be at the crate root if direct usage of the declared
969 modules within `use` items is desired. It is also possible to use `self` and
970 `super` at the beginning of a `use` item to refer to the current and direct
971 parent modules respectively. All rules regarding accessing declared modules in
972 `use` declarations apply to both module declarations and `extern crate`
975 An example of what will and will not work for `use` items:
978 # #![allow(unused_imports)]
979 use foo::baz::foobaz; // good: foo is at the root of the crate
987 use foo::example::iter; // good: foo is at crate root
988 // use example::iter; // bad: example is not at the crate root
989 use self::baz::foobaz; // good: self refers to module 'foo'
990 use foo::bar::foobar; // good: foo is at crate root
997 use super::bar::foobar; // good: super refers to module 'foo'
1007 A _function item_ defines a sequence of [statements](#statements) and a
1008 final [expression](#expressions), along with a name and a set of
1009 parameters. Other than a name, all these are optional.
1010 Functions are declared with the keyword `fn`. Functions may declare a
1011 set of *input* [*variables*](#variables) as parameters, through which the caller
1012 passes arguments into the function, and the *output* [*type*](#types)
1013 of the value the function will return to its caller on completion.
1015 A function may also be copied into a first-class *value*, in which case the
1016 value has the corresponding [*function type*](#function-types), and can be used
1017 otherwise exactly as a function item (with a minor additional cost of calling
1018 the function indirectly).
1020 Every control path in a function logically ends with a `return` expression or a
1021 diverging expression. If the outermost block of a function has a
1022 value-producing expression in its final-expression position, that expression is
1023 interpreted as an implicit `return` expression applied to the final-expression.
1025 An example of a function:
1028 fn add(x: i32, y: i32) -> i32 {
1033 As with `let` bindings, function arguments are irrefutable patterns, so any
1034 pattern that is valid in a let binding is also valid as an argument.
1037 fn first((value, _): (i32, i32)) -> i32 { value }
1041 #### Generic functions
1043 A _generic function_ allows one or more _parameterized types_ to appear in its
1044 signature. Each type parameter must be explicitly declared in an
1045 angle-bracket-enclosed and comma-separated list, following the function name.
1048 // foo is generic over A and B
1050 fn foo<A, B>(x: A, y: B) {
1053 Inside the function signature and body, the name of the type parameter can be
1054 used as a type name. [Trait](#traits) bounds can be specified for type parameters
1055 to allow methods with that trait to be called on values of that type. This is
1056 specified using the `where` syntax:
1059 fn foo<T>(x: T) where T: Debug {
1062 When a generic function is referenced, its type is instantiated based on the
1063 context of the reference. For example, calling the `foo` function here:
1066 use std::fmt::Debug;
1068 fn foo<T>(x: &[T]) where T: Debug {
1076 will instantiate type parameter `T` with `i32`.
1078 The type parameters can also be explicitly supplied in a trailing
1079 [path](#paths) component after the function name. This might be necessary if
1080 there is not sufficient context to determine the type parameters. For example,
1081 `mem::size_of::<u32>() == 4`.
1083 #### Diverging functions
1085 A special kind of function can be declared with a `!` character where the
1086 output type would normally be. For example:
1089 fn my_err(s: &str) -> ! {
1095 We call such functions "diverging" because they never return a value to the
1096 caller. Every control path in a diverging function must end with a `panic!()` or
1097 a call to another diverging function on every control path. The `!` annotation
1098 does *not* denote a type.
1100 It might be necessary to declare a diverging function because as mentioned
1101 previously, the typechecker checks that every control path in a function ends
1102 with a [`return`](#return-expressions) or diverging expression. So, if `my_err`
1103 were declared without the `!` annotation, the following code would not
1107 # fn my_err(s: &str) -> ! { panic!() }
1109 fn f(i: i32) -> i32 {
1114 my_err("Bad number!");
1119 This will not compile without the `!` annotation on `my_err`, since the `else`
1120 branch of the conditional in `f` does not return an `i32`, as required by the
1121 signature of `f`. Adding the `!` annotation to `my_err` informs the
1122 typechecker that, should control ever enter `my_err`, no further type judgments
1123 about `f` need to hold, since control will never resume in any context that
1124 relies on those judgments. Thus the return type on `f` only needs to reflect
1125 the `if` branch of the conditional.
1127 #### Extern functions
1129 Extern functions are part of Rust's foreign function interface, providing the
1130 opposite functionality to [external blocks](#external-blocks). Whereas
1131 external blocks allow Rust code to call foreign code, extern functions with
1132 bodies defined in Rust code _can be called by foreign code_. They are defined
1133 in the same way as any other Rust function, except that they have the `extern`
1137 // Declares an extern fn, the ABI defaults to "C"
1138 extern fn new_i32() -> i32 { 0 }
1140 // Declares an extern fn with "stdcall" ABI
1141 extern "stdcall" fn new_i32_stdcall() -> i32 { 0 }
1144 Unlike normal functions, extern fns have type `extern "ABI" fn()`. This is the
1145 same type as the functions declared in an extern block.
1148 # extern fn new_i32() -> i32 { 0 }
1149 let fptr: extern "C" fn() -> i32 = new_i32;
1152 Extern functions may be called directly from Rust code as Rust uses large,
1153 contiguous stack segments like C.
1157 A _type alias_ defines a new name for an existing [type](#types). Type
1158 aliases are declared with the keyword `type`. Every value has a single,
1159 specific type, but may implement several different traits, or be compatible with
1160 several different type constraints.
1162 For example, the following defines the type `Point` as a synonym for the type
1163 `(u8, u8)`, the type of pairs of unsigned 8 bit integers:
1166 type Point = (u8, u8);
1167 let p: Point = (41, 68);
1170 Currently a type alias to an enum type cannot be used to qualify the
1176 let _: F = E::A; // OK
1177 // let _: F = F::A; // Doesn't work
1182 A _struct_ is a nominal [struct type](#struct-types) defined with the
1185 An example of a `struct` item and its use:
1188 struct Point {x: i32, y: i32}
1189 let p = Point {x: 10, y: 11};
1193 A _tuple struct_ is a nominal [tuple type](#tuple-types), also defined with
1194 the keyword `struct`. For example:
1197 struct Point(i32, i32);
1198 let p = Point(10, 11);
1199 let px: i32 = match p { Point(x, _) => x };
1202 A _unit-like struct_ is a struct without any fields, defined by leaving off
1203 the list of fields entirely. Such a struct implicitly defines a constant of
1204 its type with the same name. For example:
1208 let c = [Cookie, Cookie {}, Cookie, Cookie {}];
1215 const Cookie: Cookie = Cookie {};
1216 let c = [Cookie, Cookie {}, Cookie, Cookie {}];
1219 The precise memory layout of a struct is not specified. One can specify a
1220 particular layout using the [`repr` attribute](#ffi-attributes).
1224 An _enumeration_ is a simultaneous definition of a nominal [enumerated
1225 type](#enumerated-types) as well as a set of *constructors*, that can be used
1226 to create or pattern-match values of the corresponding enumerated type.
1228 Enumerations are declared with the keyword `enum`.
1230 An example of an `enum` item and its use:
1238 let mut a: Animal = Animal::Dog;
1242 Enumeration constructors can have either named or unnamed fields:
1247 Cat { name: String, weight: f64 },
1250 let mut a: Animal = Animal::Dog("Cocoa".to_string(), 37.2);
1251 a = Animal::Cat { name: "Spotty".to_string(), weight: 2.7 };
1254 In this example, `Cat` is a _struct-like enum variant_,
1255 whereas `Dog` is simply called an enum variant.
1257 Each enum value has a _discriminant_ which is an integer associated to it. You
1258 can specify it explicitly:
1266 The right hand side of the specification is interpreted as an `isize` value,
1267 but the compiler is allowed to use a smaller type in the actual memory layout.
1268 The [`repr` attribute](#ffi-attributes) can be added in order to change
1269 the type of the right hand side and specify the memory layout.
1271 If a discriminant isn't specified, they start at zero, and add one for each
1274 You can cast an enum to get its discriminant:
1277 # enum Foo { Bar = 123 }
1278 let x = Foo::Bar as u32; // x is now 123u32
1281 This only works as long as none of the variants have data attached. If
1282 it were `Bar(i32)`, this is disallowed.
1286 A *constant item* is a named _constant value_ which is not associated with a
1287 specific memory location in the program. Constants are essentially inlined
1288 wherever they are used, meaning that they are copied directly into the relevant
1289 context when used. References to the same constant are not necessarily
1290 guaranteed to refer to the same memory address.
1292 Constant values must not have destructors, and otherwise permit most forms of
1293 data. Constants may refer to the address of other constants, in which case the
1294 address will have the `static` lifetime. The compiler is, however, still at
1295 liberty to translate the constant many times, so the address referred to may not
1298 Constants must be explicitly typed. The type may be `bool`, `char`, a number, or
1299 a type derived from those primitive types. The derived types are references with
1300 the `static` lifetime, fixed-size arrays, tuples, enum variants, and structs.
1303 const BIT1: u32 = 1 << 0;
1304 const BIT2: u32 = 1 << 1;
1306 const BITS: [u32; 2] = [BIT1, BIT2];
1307 const STRING: &'static str = "bitstring";
1309 struct BitsNStrings<'a> {
1314 const BITS_N_STRINGS: BitsNStrings<'static> = BitsNStrings {
1322 A *static item* is similar to a *constant*, except that it represents a precise
1323 memory location in the program. A static is never "inlined" at the usage site,
1324 and all references to it refer to the same memory location. Static items have
1325 the `static` lifetime, which outlives all other lifetimes in a Rust program.
1326 Static items may be placed in read-only memory if they do not contain any
1327 interior mutability.
1329 Statics may contain interior mutability through the `UnsafeCell` language item.
1330 All access to a static is safe, but there are a number of restrictions on
1333 * Statics may not contain any destructors.
1334 * The types of static values must ascribe to `Sync` to allow thread-safe access.
1335 * Statics may not refer to other statics by value, only by reference.
1336 * Constants cannot refer to statics.
1338 Constants should in general be preferred over statics, unless large amounts of
1339 data are being stored, or single-address and mutability properties are required.
1341 #### Mutable statics
1343 If a static item is declared with the `mut` keyword, then it is allowed to
1344 be modified by the program. One of Rust's goals is to make concurrency bugs
1345 hard to run into, and this is obviously a very large source of race conditions
1346 or other bugs. For this reason, an `unsafe` block is required when either
1347 reading or writing a mutable static variable. Care should be taken to ensure
1348 that modifications to a mutable static are safe with respect to other threads
1349 running in the same process.
1351 Mutable statics are still very useful, however. They can be used with C
1352 libraries and can also be bound from C libraries (in an `extern` block).
1355 # fn atomic_add(_: &mut u32, _: u32) -> u32 { 2 }
1357 static mut LEVELS: u32 = 0;
1359 // This violates the idea of no shared state, and this doesn't internally
1360 // protect against races, so this function is `unsafe`
1361 unsafe fn bump_levels_unsafe1() -> u32 {
1367 // Assuming that we have an atomic_add function which returns the old value,
1368 // this function is "safe" but the meaning of the return value may not be what
1369 // callers expect, so it's still marked as `unsafe`
1370 unsafe fn bump_levels_unsafe2() -> u32 {
1371 return atomic_add(&mut LEVELS, 1);
1375 Mutable statics have the same restrictions as normal statics, except that the
1376 type of the value is not required to ascribe to `Sync`.
1380 A _trait_ describes an abstract interface that types can
1381 implement. This interface consists of associated items, which come in
1388 Associated functions whose first parameter is named `self` are called
1389 methods and may be invoked using `.` notation (e.g., `x.foo()`).
1391 All traits define an implicit type parameter `Self` that refers to
1392 "the type that is implementing this interface". Traits may also
1393 contain additional type parameters. These type parameters (including
1394 `Self`) may be constrained by other traits and so forth as usual.
1396 Trait bounds on `Self` are considered "supertraits". These are
1397 required to be acyclic. Supertraits are somewhat different from other
1398 constraints in that they affect what methods are available in the
1399 vtable when the trait is used as a [trait object](#trait-objects).
1401 Traits are implemented for specific types through separate
1402 [implementations](#implementations).
1404 Consider the following trait:
1407 # type Surface = i32;
1408 # type BoundingBox = i32;
1410 fn draw(&self, Surface);
1411 fn bounding_box(&self) -> BoundingBox;
1415 This defines a trait with two methods. All values that have
1416 [implementations](#implementations) of this trait in scope can have their
1417 `draw` and `bounding_box` methods called, using `value.bounding_box()`
1418 [syntax](#method-call-expressions).
1420 Traits can include default implementations of methods, as in:
1425 fn baz(&self) { println!("We called baz."); }
1429 Here the `baz` method has a default implementation, so types that implement
1430 `Foo` need only implement `bar`. It is also possible for implementing types
1431 to override a method that has a default implementation.
1433 Type parameters can be specified for a trait to make it generic. These appear
1434 after the trait name, using the same syntax used in [generic
1435 functions](#generic-functions).
1439 fn len(&self) -> u32;
1440 fn elt_at(&self, n: u32) -> T;
1441 fn iter<F>(&self, F) where F: Fn(T);
1445 It is also possible to define associated types for a trait. Consider the
1446 following example of a `Container` trait. Notice how the type is available
1447 for use in the method signatures:
1453 fn insert(&mut self, Self::E);
1457 In order for a type to implement this trait, it must not only provide
1458 implementations for every method, but it must specify the type `E`. Here's
1459 an implementation of `Container` for the standard library type `Vec`:
1464 # fn empty() -> Self;
1465 # fn insert(&mut self, Self::E);
1467 impl<T> Container for Vec<T> {
1469 fn empty() -> Vec<T> { Vec::new() }
1470 fn insert(&mut self, x: T) { self.push(x); }
1474 Generic functions may use traits as _bounds_ on their type parameters. This
1475 will have two effects:
1477 - Only types that have the trait may instantiate the parameter.
1478 - Within the generic function, the methods of the trait can be
1479 called on values that have the parameter's type.
1484 # type Surface = i32;
1485 # trait Shape { fn draw(&self, Surface); }
1486 fn draw_twice<T: Shape>(surface: Surface, sh: T) {
1492 Traits also define a [trait object](#trait-objects) with the same
1493 name as the trait. Values of this type are created by coercing from a
1494 pointer of some specific type to a pointer of trait type. For example,
1495 `&T` could be coerced to `&Shape` if `T: Shape` holds (and similarly
1496 for `Box<T>`). This coercion can either be implicit or
1497 [explicit](#type-cast-expressions). Here is an example of an explicit
1502 impl Shape for i32 { }
1503 let mycircle = 0i32;
1504 let myshape: Box<Shape> = Box::new(mycircle) as Box<Shape>;
1507 The resulting value is a box containing the value that was cast, along with
1508 information that identifies the methods of the implementation that was used.
1509 Values with a trait type can have [methods called](#method-call-expressions) on
1510 them, for any method in the trait, and can be used to instantiate type
1511 parameters that are bounded by the trait.
1513 Trait methods may be static, which means that they lack a `self` argument.
1514 This means that they can only be called with function call syntax (`f(x)`) and
1515 not method call syntax (`obj.f()`). The way to refer to the name of a static
1516 method is to qualify it with the trait name, treating the trait name like a
1517 module. For example:
1521 fn from_i32(n: i32) -> Self;
1524 fn from_i32(n: i32) -> f64 { n as f64 }
1526 let x: f64 = Num::from_i32(42);
1529 Traits may inherit from other traits. Consider the following example:
1532 trait Shape { fn area(&self) -> f64; }
1533 trait Circle : Shape { fn radius(&self) -> f64; }
1536 The syntax `Circle : Shape` means that types that implement `Circle` must also
1537 have an implementation for `Shape`. Multiple supertraits are separated by `+`,
1538 `trait Circle : Shape + PartialEq { }`. In an implementation of `Circle` for a
1539 given type `T`, methods can refer to `Shape` methods, since the typechecker
1540 checks that any type with an implementation of `Circle` also has an
1541 implementation of `Shape`:
1546 trait Shape { fn area(&self) -> f64; }
1547 trait Circle : Shape { fn radius(&self) -> f64; }
1548 impl Shape for Foo {
1549 fn area(&self) -> f64 {
1553 impl Circle for Foo {
1554 fn radius(&self) -> f64 {
1555 println!("calling area: {}", self.area());
1565 In type-parameterized functions, methods of the supertrait may be called on
1566 values of subtrait-bound type parameters. Referring to the previous example of
1567 `trait Circle : Shape`:
1570 # trait Shape { fn area(&self) -> f64; }
1571 # trait Circle : Shape { fn radius(&self) -> f64; }
1572 fn radius_times_area<T: Circle>(c: T) -> f64 {
1573 // `c` is both a Circle and a Shape
1574 c.radius() * c.area()
1578 Likewise, supertrait methods may also be called on trait objects.
1581 # trait Shape { fn area(&self) -> f64; }
1582 # trait Circle : Shape { fn radius(&self) -> f64; }
1583 # impl Shape for i32 { fn area(&self) -> f64 { 0.0 } }
1584 # impl Circle for i32 { fn radius(&self) -> f64 { 0.0 } }
1585 # let mycircle = 0i32;
1586 let mycircle = Box::new(mycircle) as Box<Circle>;
1587 let nonsense = mycircle.radius() * mycircle.area();
1592 An _implementation_ is an item that implements a [trait](#traits) for a
1595 Implementations are defined with the keyword `impl`.
1598 # #[derive(Copy, Clone)]
1599 # struct Point {x: f64, y: f64};
1600 # type Surface = i32;
1601 # struct BoundingBox {x: f64, y: f64, width: f64, height: f64};
1602 # trait Shape { fn draw(&self, Surface); fn bounding_box(&self) -> BoundingBox; }
1603 # fn do_draw_circle(s: Surface, c: Circle) { }
1609 impl Copy for Circle {}
1611 impl Clone for Circle {
1612 fn clone(&self) -> Circle { *self }
1615 impl Shape for Circle {
1616 fn draw(&self, s: Surface) { do_draw_circle(s, *self); }
1617 fn bounding_box(&self) -> BoundingBox {
1618 let r = self.radius;
1620 x: self.center.x - r,
1621 y: self.center.y - r,
1629 It is possible to define an implementation without referring to a trait. The
1630 methods in such an implementation can only be used as direct calls on the values
1631 of the type that the implementation targets. In such an implementation, the
1632 trait type and `for` after `impl` are omitted. Such implementations are limited
1633 to nominal types (enums, structs, trait objects), and the implementation must
1634 appear in the same crate as the `self` type:
1637 struct Point {x: i32, y: i32}
1641 println!("Point is at ({}, {})", self.x, self.y);
1645 let my_point = Point {x: 10, y:11};
1649 When a trait _is_ specified in an `impl`, all methods declared as part of the
1650 trait must be implemented, with matching types and type parameter counts.
1652 An implementation can take type parameters, which can be different from the
1653 type parameters taken by the trait it implements. Implementation parameters
1654 are written after the `impl` keyword.
1657 # trait Seq<T> { fn dummy(&self, _: T) { } }
1658 impl<T> Seq<T> for Vec<T> {
1661 impl Seq<bool> for u32 {
1662 /* Treat the integer as a sequence of bits */
1668 External blocks form the basis for Rust's foreign function interface.
1669 Declarations in an external block describe symbols in external, non-Rust
1672 Functions within external blocks are declared in the same way as other Rust
1673 functions, with the exception that they may not have a body and are instead
1674 terminated by a semicolon.
1676 Functions within external blocks may be called by Rust code, just like
1677 functions defined in Rust. The Rust compiler automatically translates between
1678 the Rust ABI and the foreign ABI.
1680 Functions within external blocks may be variadic by specifying `...` after one
1681 or more named arguments in the argument list:
1685 fn foo(x: i32, ...);
1689 A number of [attributes](#ffi-attributes) control the behavior of external blocks.
1691 By default external blocks assume that the library they are calling uses the
1692 standard C ABI on the specific platform. Other ABIs may be specified using an
1693 `abi` string, as shown here:
1696 // Interface to the Windows API
1697 extern "stdcall" { }
1700 There are three ABI strings which are cross-platform, and which all compilers
1701 are guaranteed to support:
1703 * `extern "Rust"` -- The default ABI when you write a normal `fn foo()` in any
1705 * `extern "C"` -- This is the same as `extern fn foo()`; whatever the default
1706 your C compiler supports.
1707 * `extern "system"` -- Usually the same as `extern "C"`, except on Win32, in
1708 which case it's `"stdcall"`, or what you should use to link to the Windows API
1711 There are also some platform-specific ABI strings:
1713 * `extern "cdecl"` -- The default for x86\_32 C code.
1714 * `extern "stdcall"` -- The default for the Win32 API on x86\_32.
1715 * `extern "win64"` -- The default for C code on x86\_64 Windows.
1716 * `extern "sysv64"` -- The default for C code on non-Windows x86\_64.
1717 * `extern "aapcs"` -- The default for ARM.
1718 * `extern "fastcall"` -- The `fastcall` ABI -- corresponds to MSVC's
1719 `__fastcall` and GCC and clang's `__attribute__((fastcall))`
1720 * `extern "vectorcall"` -- The `vectorcall` ABI -- corresponds to MSVC's
1721 `__vectorcall` and clang's `__attribute__((vectorcall))`
1723 Finally, there are some rustc-specific ABI strings:
1725 * `extern "rust-intrinsic"` -- The ABI of rustc intrinsics.
1726 * `extern "rust-call"` -- The ABI of the Fn::call trait functions.
1727 * `extern "platform-intrinsic"` -- Specific platform intrinsics -- like, for
1728 example, `sqrt` -- have this ABI. You should never have to deal with it.
1730 The `link` attribute allows the name of the library to be specified. When
1731 specified the compiler will attempt to link against the native library of the
1735 #[link(name = "crypto")]
1739 The type of a function declared in an extern block is `extern "abi" fn(A1, ...,
1740 An) -> R`, where `A1...An` are the declared types of its arguments and `R` is
1741 the declared return type.
1743 It is valid to add the `link` attribute on an empty extern block. You can use
1744 this to satisfy the linking requirements of extern blocks elsewhere in your code
1745 (including upstream crates) instead of adding the attribute to each extern block.
1747 ## Visibility and Privacy
1749 These two terms are often used interchangeably, and what they are attempting to
1750 convey is the answer to the question "Can this item be used at this location?"
1752 Rust's name resolution operates on a global hierarchy of namespaces. Each level
1753 in the hierarchy can be thought of as some item. The items are one of those
1754 mentioned above, but also include external crates. Declaring or defining a new
1755 module can be thought of as inserting a new tree into the hierarchy at the
1756 location of the definition.
1758 To control whether interfaces can be used across modules, Rust checks each use
1759 of an item to see whether it should be allowed or not. This is where privacy
1760 warnings are generated, or otherwise "you used a private item of another module
1761 and weren't allowed to."
1763 By default, everything in Rust is *private*, with two exceptions: Associated
1764 items in a `pub` Trait are public by default; Enum variants
1765 in a `pub` enum are also public by default. When an item is declared as `pub`,
1766 it can be thought of as being accessible to the outside world. For example:
1770 // Declare a private struct
1773 // Declare a public struct with a private field
1778 // Declare a public enum with two public variants
1780 PubliclyAccessibleState,
1781 PubliclyAccessibleState2,
1785 With the notion of an item being either public or private, Rust allows item
1786 accesses in two cases:
1788 1. If an item is public, then it can be used externally through any of its
1790 2. If an item is private, it may be accessed by the current module and its
1793 These two cases are surprisingly powerful for creating module hierarchies
1794 exposing public APIs while hiding internal implementation details. To help
1795 explain, here's a few use cases and what they would entail:
1797 * A library developer needs to expose functionality to crates which link
1798 against their library. As a consequence of the first case, this means that
1799 anything which is usable externally must be `pub` from the root down to the
1800 destination item. Any private item in the chain will disallow external
1803 * A crate needs a global available "helper module" to itself, but it doesn't
1804 want to expose the helper module as a public API. To accomplish this, the
1805 root of the crate's hierarchy would have a private module which then
1806 internally has a "public API". Because the entire crate is a descendant of
1807 the root, then the entire local crate can access this private module through
1810 * When writing unit tests for a module, it's often a common idiom to have an
1811 immediate child of the module to-be-tested named `mod test`. This module
1812 could access any items of the parent module through the second case, meaning
1813 that internal implementation details could also be seamlessly tested from the
1816 In the second case, it mentions that a private item "can be accessed" by the
1817 current module and its descendants, but the exact meaning of accessing an item
1818 depends on what the item is. Accessing a module, for example, would mean
1819 looking inside of it (to import more items). On the other hand, accessing a
1820 function would mean that it is invoked. Additionally, path expressions and
1821 import statements are considered to access an item in the sense that the
1822 import/expression is only valid if the destination is in the current visibility
1825 Here's an example of a program which exemplifies the three cases outlined
1829 // This module is private, meaning that no external crate can access this
1830 // module. Because it is private at the root of this current crate, however, any
1831 // module in the crate may access any publicly visible item in this module.
1832 mod crate_helper_module {
1834 // This function can be used by anything in the current crate
1835 pub fn crate_helper() {}
1837 // This function *cannot* be used by anything else in the crate. It is not
1838 // publicly visible outside of the `crate_helper_module`, so only this
1839 // current module and its descendants may access it.
1840 fn implementation_detail() {}
1843 // This function is "public to the root" meaning that it's available to external
1844 // crates linking against this one.
1845 pub fn public_api() {}
1847 // Similarly to 'public_api', this module is public so external crates may look
1850 use crate_helper_module;
1852 pub fn my_method() {
1853 // Any item in the local crate may invoke the helper module's public
1854 // interface through a combination of the two rules above.
1855 crate_helper_module::crate_helper();
1858 // This function is hidden to any module which is not a descendant of
1860 fn my_implementation() {}
1866 fn test_my_implementation() {
1867 // Because this module is a descendant of `submodule`, it's allowed
1868 // to access private items inside of `submodule` without a privacy
1870 super::my_implementation();
1878 For a Rust program to pass the privacy checking pass, all paths must be valid
1879 accesses given the two rules above. This includes all use statements,
1880 expressions, types, etc.
1882 ### Re-exporting and Visibility
1884 Rust allows publicly re-exporting items through a `pub use` directive. Because
1885 this is a public directive, this allows the item to be used in the current
1886 module through the rules above. It essentially allows public access into the
1887 re-exported item. For example, this program is valid:
1890 pub use self::implementation::api;
1892 mod implementation {
1901 This means that any external crate referencing `implementation::api::f` would
1902 receive a privacy violation, while the path `api::f` would be allowed.
1904 When re-exporting a private item, it can be thought of as allowing the "privacy
1905 chain" being short-circuited through the reexport instead of passing through
1906 the namespace hierarchy as it normally would.
1910 Any item declaration may have an _attribute_ applied to it. Attributes in Rust
1911 are modeled on Attributes in ECMA-335, with the syntax coming from ECMA-334
1912 (C#). An attribute is a general, free-form metadatum that is interpreted
1913 according to name, convention, and language and compiler version. Attributes
1914 may appear as any of:
1916 * A single identifier, the attribute name
1917 * An identifier followed by the equals sign '=' and a literal, providing a
1919 * An identifier followed by a parenthesized list of sub-attribute arguments
1921 Attributes with a bang ("!") after the hash ("#") apply to the item that the
1922 attribute is declared within. Attributes that do not have a bang after the hash
1923 apply to the item that follows the attribute.
1925 An example of attributes:
1928 // General metadata applied to the enclosing module or crate.
1929 #![crate_type = "lib"]
1931 // A function marked as a unit test
1937 // A conditionally-compiled module
1938 #[cfg(target_os="linux")]
1943 // A lint attribute used to suppress a warning/error
1944 #[allow(non_camel_case_types)]
1948 > **Note:** At some point in the future, the compiler will distinguish between
1949 > language-reserved and user-available attributes. Until then, there is
1950 > effectively no difference between an attribute handled by a loadable syntax
1951 > extension and the compiler.
1953 ### Crate-only attributes
1955 - `crate_name` - specify the crate's crate name.
1956 - `crate_type` - see [linkage](#linkage).
1957 - `feature` - see [compiler features](#compiler-features).
1958 - `no_builtins` - disable optimizing certain code patterns to invocations of
1959 library functions that are assumed to exist
1960 - `no_main` - disable emitting the `main` symbol. Useful when some other
1961 object being linked to defines `main`.
1962 - `no_start` - disable linking to the `native` crate, which specifies the
1963 "start" language item.
1964 - `no_std` - disable linking to the `std` crate.
1965 - `plugin` - load a list of named crates as compiler plugins, e.g.
1966 `#![plugin(foo, bar)]`. Optional arguments for each plugin,
1967 i.e. `#![plugin(foo(... args ...))]`, are provided to the plugin's
1968 registrar function. The `plugin` feature gate is required to use
1970 - `recursion_limit` - Sets the maximum depth for potentially
1971 infinitely-recursive compile-time operations like
1972 auto-dereference or macro expansion. The default is
1973 `#![recursion_limit="64"]`.
1975 ### Module-only attributes
1977 - `no_implicit_prelude` - disable injecting `use std::prelude::*` in this
1979 - `path` - specifies the file to load the module from. `#[path="foo.rs"] mod
1980 bar;` is equivalent to `mod bar { /* contents of foo.rs */ }`. The path is
1981 taken relative to the directory that the current module is in.
1983 ### Function-only attributes
1985 - `main` - indicates that this function should be passed to the entry point,
1986 rather than the function in the crate root named `main`.
1987 - `plugin_registrar` - mark this function as the registration point for
1988 [compiler plugins][plugin], such as loadable syntax extensions.
1989 - `start` - indicates that this function should be used as the entry point,
1990 overriding the "start" language item. See the "start" [language
1991 item](#language-items) for more details.
1992 - `test` - indicates that this function is a test function, to only be compiled
1993 in case of `--test`.
1994 - `should_panic` - indicates that this test function should panic, inverting the success condition.
1995 - `cold` - The function is unlikely to be executed, so optimize it (and calls
1997 - `naked` - The function utilizes a custom ABI or custom inline ASM that requires
1998 epilogue and prologue to be skipped.
2000 ### Static-only attributes
2002 - `thread_local` - on a `static mut`, this signals that the value of this
2003 static may change depending on the current thread. The exact consequences of
2004 this are implementation-defined.
2008 On an `extern` block, the following attributes are interpreted:
2010 - `link_args` - specify arguments to the linker, rather than just the library
2011 name and type. This is feature gated and the exact behavior is
2012 implementation-defined (due to variety of linker invocation syntax).
2013 - `link` - indicate that a native library should be linked to for the
2014 declarations in this block to be linked correctly. `link` supports an optional
2015 `kind` key with three possible values: `dylib`, `static`, and `framework`. See
2016 [external blocks](#external-blocks) for more about external blocks. Two
2017 examples: `#[link(name = "readline")]` and
2018 `#[link(name = "CoreFoundation", kind = "framework")]`.
2019 - `linked_from` - indicates what native library this block of FFI items is
2020 coming from. This attribute is of the form `#[linked_from = "foo"]` where
2021 `foo` is the name of a library in either `#[link]` or a `-l` flag. This
2022 attribute is currently required to export symbols from a Rust dynamic library
2023 on Windows, and it is feature gated behind the `linked_from` feature.
2025 On declarations inside an `extern` block, the following attributes are
2028 - `link_name` - the name of the symbol that this function or static should be
2030 - `linkage` - on a static, this specifies the [linkage
2031 type](http://llvm.org/docs/LangRef.html#linkage-types).
2035 - `repr` - on C-like enums, this sets the underlying type used for
2036 representation. Takes one argument, which is the primitive
2037 type this enum should be represented for, or `C`, which specifies that it
2038 should be the default `enum` size of the C ABI for that platform. Note that
2039 enum representation in C is undefined, and this may be incorrect when the C
2040 code is compiled with certain flags.
2044 - `repr` - specifies the representation to use for this struct. Takes a list
2045 of options. The currently accepted ones are `C` and `packed`, which may be
2046 combined. `C` will use a C ABI compatible struct layout, and `packed` will
2047 remove any padding between fields (note that this is very fragile and may
2048 break platforms which require aligned access).
2050 ### Macro-related attributes
2052 - `macro_use` on a `mod` — macros defined in this module will be visible in the
2053 module's parent, after this module has been included.
2055 - `macro_use` on an `extern crate` — load macros from this crate. An optional
2056 list of names `#[macro_use(foo, bar)]` restricts the import to just those
2057 macros named. The `extern crate` must appear at the crate root, not inside
2058 `mod`, which ensures proper function of the [`$crate` macro
2059 variable](book/macros.html#the-variable-crate).
2061 - `macro_reexport` on an `extern crate` — re-export the named macros.
2063 - `macro_export` - export a macro for cross-crate usage.
2065 - `no_link` on an `extern crate` — even if we load this crate for macros, don't
2066 link it into the output.
2068 See the [macros section of the
2069 book](book/macros.html#scoping-and-macro-importexport) for more information on
2073 ### Miscellaneous attributes
2075 - `deprecated` - mark the item as deprecated; the full attribute is `#[deprecated(since = "crate version", note = "...")`, where both arguments are optional.
2076 - `export_name` - on statics and functions, this determines the name of the
2078 - `link_section` - on statics and functions, this specifies the section of the
2079 object file that this item's contents will be placed into.
2080 - `no_mangle` - on any item, do not apply the standard name mangling. Set the
2081 symbol for this item to its identifier.
2082 - `simd` - on certain tuple structs, derive the arithmetic operators, which
2083 lower to the target's SIMD instructions, if any; the `simd` feature gate
2084 is necessary to use this attribute.
2085 - `unsafe_destructor_blind_to_params` - on `Drop::drop` method, asserts that the
2086 destructor code (and all potential specializations of that code) will
2087 never attempt to read from nor write to any references with lifetimes
2088 that come in via generic parameters. This is a constraint we cannot
2089 currently express via the type system, and therefore we rely on the
2090 programmer to assert that it holds. Adding this to a Drop impl causes
2091 the associated destructor to be considered "uninteresting" by the
2092 Drop-Check rule, and thus it can help sidestep data ordering
2093 constraints that would otherwise be introduced by the Drop-Check
2094 rule. Such sidestepping of the constraints, if done incorrectly, can
2095 lead to undefined behavior (in the form of reading or writing to data
2096 outside of its dynamic extent), and thus this attribute has the word
2097 "unsafe" in its name. To use this, the
2098 `unsafe_destructor_blind_to_params` feature gate must be enabled.
2099 - `doc` - Doc comments such as `/// foo` are equivalent to `#[doc = "foo"]`.
2100 - `rustc_on_unimplemented` - Write a custom note to be shown along with the error
2101 when the trait is found to be unimplemented on a type.
2102 You may use format arguments like `{T}`, `{A}` to correspond to the
2103 types at the point of use corresponding to the type parameters of the
2104 trait of the same name. `{Self}` will be replaced with the type that is supposed
2105 to implement the trait but doesn't. To use this, the `on_unimplemented` feature gate
2107 - `must_use` - on structs and enums, will warn if a value of this type isn't used or
2108 assigned to a variable. You may also include an optional message by using
2109 `#[must_use = "message"]` which will be given alongside the warning.
2111 ### Conditional compilation
2113 Sometimes one wants to have different compiler outputs from the same code,
2114 depending on build target, such as targeted operating system, or to enable
2117 There are two kinds of configuration options, one that is either defined or not
2118 (`#[cfg(foo)]`), and the other that contains a string that can be checked
2119 against (`#[cfg(bar = "baz")]`). Currently, only compiler-defined configuration
2120 options can have the latter form.
2123 // The function is only included in the build when compiling for OSX
2124 #[cfg(target_os = "macos")]
2129 // This function is only included when either foo or bar is defined
2130 #[cfg(any(foo, bar))]
2131 fn needs_foo_or_bar() {
2135 // This function is only included when compiling for a unixish OS with a 32-bit
2137 #[cfg(all(unix, target_pointer_width = "32"))]
2138 fn on_32bit_unix() {
2142 // This function is only included when foo is not defined
2144 fn needs_not_foo() {
2149 This illustrates some conditional compilation can be achieved using the
2150 `#[cfg(...)]` attribute. `any`, `all` and `not` can be used to assemble
2151 arbitrarily complex configurations through nesting.
2153 The following configurations must be defined by the implementation:
2155 * `target_arch = "..."` - Target CPU architecture, such as `"x86"`,
2156 `"x86_64"` `"mips"`, `"powerpc"`, `"powerpc64"`, `"arm"`, or
2157 `"aarch64"`. This value is closely related to the first element of
2158 the platform target triple, though it is not identical.
2159 * `target_os = "..."` - Operating system of the target, examples
2160 include `"windows"`, `"macos"`, `"ios"`, `"linux"`, `"android"`,
2161 `"freebsd"`, `"dragonfly"`, `"bitrig"` , `"openbsd"` or
2162 `"netbsd"`. This value is closely related to the second and third
2163 element of the platform target triple, though it is not identical.
2164 * `target_family = "..."` - Operating system family of the target, e. g.
2165 `"unix"` or `"windows"`. The value of this configuration option is defined
2166 as a configuration itself, like `unix` or `windows`.
2167 * `unix` - See `target_family`.
2168 * `windows` - See `target_family`.
2169 * `target_env = ".."` - Further disambiguates the target platform with
2170 information about the ABI/libc. Presently this value is either
2171 `"gnu"`, `"msvc"`, `"musl"`, or the empty string. For historical
2172 reasons this value has only been defined as non-empty when needed
2173 for disambiguation. Thus on many GNU platforms this value will be
2174 empty. This value is closely related to the fourth element of the
2175 platform target triple, though it is not identical. For example,
2176 embedded ABIs such as `gnueabihf` will simply define `target_env` as
2178 * `target_endian = "..."` - Endianness of the target CPU, either `"little"` or
2180 * `target_pointer_width = "..."` - Target pointer width in bits. This is set
2181 to `"32"` for targets with 32-bit pointers, and likewise set to `"64"` for
2183 * `target_has_atomic = "..."` - Set of integer sizes on which the target can perform
2184 atomic operations. Values are `"8"`, `"16"`, `"32"`, `"64"` and `"ptr"`.
2185 * `target_vendor = "..."` - Vendor of the target, for example `apple`, `pc`, or
2187 * `test` - Enabled when compiling the test harness (using the `--test` flag).
2188 * `debug_assertions` - Enabled by default when compiling without optimizations.
2189 This can be used to enable extra debugging code in development but not in
2190 production. For example, it controls the behavior of the standard library's
2191 `debug_assert!` macro.
2193 You can also set another attribute based on a `cfg` variable with `cfg_attr`:
2199 Will be the same as `#[b]` if `a` is set by `cfg`, and nothing otherwise.
2201 ### Lint check attributes
2203 A lint check names a potentially undesirable coding pattern, such as
2204 unreachable code or omitted documentation, for the static entity to which the
2207 For any lint check `C`:
2209 * `allow(C)` overrides the check for `C` so that violations will go
2211 * `deny(C)` signals an error after encountering a violation of `C`,
2212 * `forbid(C)` is the same as `deny(C)`, but also forbids changing the lint
2214 * `warn(C)` warns about violations of `C` but continues compilation.
2216 The lint checks supported by the compiler can be found via `rustc -W help`,
2217 along with their default settings. [Compiler
2218 plugins](book/compiler-plugins.html#lint-plugins) can provide additional lint checks.
2222 // Missing documentation is ignored here
2223 #[allow(missing_docs)]
2224 pub fn undocumented_one() -> i32 { 1 }
2226 // Missing documentation signals a warning here
2227 #[warn(missing_docs)]
2228 pub fn undocumented_too() -> i32 { 2 }
2230 // Missing documentation signals an error here
2231 #[deny(missing_docs)]
2232 pub fn undocumented_end() -> i32 { 3 }
2236 This example shows how one can use `allow` and `warn` to toggle a particular
2240 #[warn(missing_docs)]
2242 #[allow(missing_docs)]
2244 // Missing documentation is ignored here
2245 pub fn undocumented_one() -> i32 { 1 }
2247 // Missing documentation signals a warning here,
2248 // despite the allow above.
2249 #[warn(missing_docs)]
2250 pub fn undocumented_two() -> i32 { 2 }
2253 // Missing documentation signals a warning here
2254 pub fn undocumented_too() -> i32 { 3 }
2258 This example shows how one can use `forbid` to disallow uses of `allow` for
2262 #[forbid(missing_docs)]
2264 // Attempting to toggle warning signals an error here
2265 #[allow(missing_docs)]
2267 pub fn undocumented_too() -> i32 { 2 }
2273 Some primitive Rust operations are defined in Rust code, rather than being
2274 implemented directly in C or assembly language. The definitions of these
2275 operations have to be easy for the compiler to find. The `lang` attribute
2276 makes it possible to declare these operations. For example, the `str` module
2277 in the Rust standard library defines the string equality function:
2281 pub fn eq_slice(a: &str, b: &str) -> bool {
2286 The name `str_eq` has a special meaning to the Rust compiler, and the presence
2287 of this definition means that it will use this definition when generating calls
2288 to the string equality function.
2290 The set of language items is currently considered unstable. A complete
2291 list of the built-in language items will be added in the future.
2293 ### Inline attributes
2295 The inline attribute suggests that the compiler should place a copy of
2296 the function or static in the caller, rather than generating code to
2297 call the function or access the static where it is defined.
2299 The compiler automatically inlines functions based on internal heuristics.
2300 Incorrectly inlining functions can actually make the program slower, so it
2301 should be used with care.
2303 `#[inline]` and `#[inline(always)]` always cause the function to be serialized
2304 into the crate metadata to allow cross-crate inlining.
2306 There are three different types of inline attributes:
2308 * `#[inline]` hints the compiler to perform an inline expansion.
2309 * `#[inline(always)]` asks the compiler to always perform an inline expansion.
2310 * `#[inline(never)]` asks the compiler to never perform an inline expansion.
2314 The `derive` attribute allows certain traits to be automatically implemented
2315 for data structures. For example, the following will create an `impl` for the
2316 `PartialEq` and `Clone` traits for `Foo`, the type parameter `T` will be given
2317 the `PartialEq` or `Clone` constraints for the appropriate `impl`:
2320 #[derive(PartialEq, Clone)]
2327 The generated `impl` for `PartialEq` is equivalent to
2330 # struct Foo<T> { a: i32, b: T }
2331 impl<T: PartialEq> PartialEq for Foo<T> {
2332 fn eq(&self, other: &Foo<T>) -> bool {
2333 self.a == other.a && self.b == other.b
2336 fn ne(&self, other: &Foo<T>) -> bool {
2337 self.a != other.a || self.b != other.b
2342 You can implement `derive` for your own type through [procedural
2343 macros](#procedural-macros).
2345 ### Compiler Features
2347 Certain aspects of Rust may be implemented in the compiler, but they're not
2348 necessarily ready for every-day use. These features are often of "prototype
2349 quality" or "almost production ready", but may not be stable enough to be
2350 considered a full-fledged language feature.
2352 For this reason, Rust recognizes a special crate-level attribute of the form:
2355 #![feature(feature1, feature2, feature3)]
2358 This directive informs the compiler that the feature list: `feature1`,
2359 `feature2`, and `feature3` should all be enabled. This is only recognized at a
2360 crate-level, not at a module-level. Without this directive, all features are
2361 considered off, and using the features will result in a compiler error.
2363 The currently implemented features of the reference compiler are:
2365 * `advanced_slice_patterns` - See the [match expressions](#match-expressions)
2366 section for discussion; the exact semantics of
2367 slice patterns are subject to change, so some types
2370 * `slice_patterns` - OK, actually, slice patterns are just scary and
2371 completely unstable.
2373 * `asm` - The `asm!` macro provides a means for inline assembly. This is often
2374 useful, but the exact syntax for this feature along with its
2375 semantics are likely to change, so this macro usage must be opted
2378 * `associated_consts` - Allows constants to be defined in `impl` and `trait`
2379 blocks, so that they can be associated with a type or
2380 trait in a similar manner to methods and associated
2383 * `box_patterns` - Allows `box` patterns, the exact semantics of which
2384 is subject to change.
2386 * `box_syntax` - Allows use of `box` expressions, the exact semantics of which
2387 is subject to change.
2389 * `cfg_target_vendor` - Allows conditional compilation using the `target_vendor`
2390 matcher which is subject to change.
2392 * `cfg_target_has_atomic` - Allows conditional compilation using the `target_has_atomic`
2393 matcher which is subject to change.
2395 * `concat_idents` - Allows use of the `concat_idents` macro, which is in many
2396 ways insufficient for concatenating identifiers, and may be
2397 removed entirely for something more wholesome.
2399 * `custom_attribute` - Allows the usage of attributes unknown to the compiler
2400 so that new attributes can be added in a backwards compatible
2403 * `custom_derive` - Allows the use of `#[derive(Foo,Bar)]` as sugar for
2404 `#[derive_Foo] #[derive_Bar]`, which can be user-defined syntax
2407 * `inclusive_range_syntax` - Allows use of the `a...b` and `...b` syntax for inclusive ranges.
2409 * `inclusive_range` - Allows use of the types that represent desugared inclusive ranges.
2411 * `intrinsics` - Allows use of the "rust-intrinsics" ABI. Compiler intrinsics
2412 are inherently unstable and no promise about them is made.
2414 * `lang_items` - Allows use of the `#[lang]` attribute. Like `intrinsics`,
2415 lang items are inherently unstable and no promise about them
2418 * `link_args` - This attribute is used to specify custom flags to the linker,
2419 but usage is strongly discouraged. The compiler's usage of the
2420 system linker is not guaranteed to continue in the future, and
2421 if the system linker is not used then specifying custom flags
2422 doesn't have much meaning.
2424 * `link_llvm_intrinsics` – Allows linking to LLVM intrinsics via
2425 `#[link_name="llvm.*"]`.
2427 * `linkage` - Allows use of the `linkage` attribute, which is not portable.
2429 * `log_syntax` - Allows use of the `log_syntax` macro attribute, which is a
2430 nasty hack that will certainly be removed.
2432 * `main` - Allows use of the `#[main]` attribute, which changes the entry point
2433 into a Rust program. This capability is subject to change.
2435 * `macro_reexport` - Allows macros to be re-exported from one crate after being imported
2436 from another. This feature was originally designed with the sole
2437 use case of the Rust standard library in mind, and is subject to
2440 * `non_ascii_idents` - The compiler supports the use of non-ascii identifiers,
2441 but the implementation is a little rough around the
2442 edges, so this can be seen as an experimental feature
2443 for now until the specification of identifiers is fully
2446 * `no_std` - Allows the `#![no_std]` crate attribute, which disables the implicit
2447 `extern crate std`. This typically requires use of the unstable APIs
2448 behind the libstd "facade", such as libcore and libcollections. It
2449 may also cause problems when using syntax extensions, including
2452 * `on_unimplemented` - Allows the `#[rustc_on_unimplemented]` attribute, which allows
2453 trait definitions to add specialized notes to error messages
2454 when an implementation was expected but not found.
2456 * `optin_builtin_traits` - Allows the definition of default and negative trait
2457 implementations. Experimental.
2459 * `plugin` - Usage of [compiler plugins][plugin] for custom lints or syntax extensions.
2460 These depend on compiler internals and are subject to change.
2462 * `plugin_registrar` - Indicates that a crate provides [compiler plugins][plugin].
2464 * `quote` - Allows use of the `quote_*!` family of macros, which are
2465 implemented very poorly and will likely change significantly
2466 with a proper implementation.
2468 * `rustc_attrs` - Gates internal `#[rustc_*]` attributes which may be
2469 for internal use only or have meaning added to them in the future.
2471 * `rustc_diagnostic_macros`- A mysterious feature, used in the implementation
2472 of rustc, not meant for mortals.
2474 * `simd` - Allows use of the `#[simd]` attribute, which is overly simple and
2475 not the SIMD interface we want to expose in the long term.
2477 * `simd_ffi` - Allows use of SIMD vectors in signatures for foreign functions.
2478 The SIMD interface is subject to change.
2480 * `start` - Allows use of the `#[start]` attribute, which changes the entry point
2481 into a Rust program. This capability, especially the signature for the
2482 annotated function, is subject to change.
2484 * `static_in_const` - Enables lifetime elision with a `'static` default for
2485 `const` and `static` item declarations.
2487 * `thread_local` - The usage of the `#[thread_local]` attribute is experimental
2488 and should be seen as unstable. This attribute is used to
2489 declare a `static` as being unique per-thread leveraging
2490 LLVM's implementation which works in concert with the kernel
2491 loader and dynamic linker. This is not necessarily available
2492 on all platforms, and usage of it is discouraged.
2494 * `trace_macros` - Allows use of the `trace_macros` macro, which is a nasty
2495 hack that will certainly be removed.
2497 * `unboxed_closures` - Rust's new closure design, which is currently a work in
2498 progress feature with many known bugs.
2500 * `allow_internal_unstable` - Allows `macro_rules!` macros to be tagged with the
2501 `#[allow_internal_unstable]` attribute, designed
2502 to allow `std` macros to call
2503 `#[unstable]`/feature-gated functionality
2504 internally without imposing on callers
2505 (i.e. making them behave like function calls in
2506 terms of encapsulation).
2508 * `default_type_parameter_fallback` - Allows type parameter defaults to
2509 influence type inference.
2511 * `stmt_expr_attributes` - Allows attributes on expressions.
2513 * `type_ascription` - Allows type ascription expressions `expr: Type`.
2515 * `abi_vectorcall` - Allows the usage of the vectorcall calling convention
2516 (e.g. `extern "vectorcall" func fn_();`)
2518 * `abi_sysv64` - Allows the usage of the system V AMD64 calling convention
2519 (e.g. `extern "sysv64" func fn_();`)
2521 If a feature is promoted to a language feature, then all existing programs will
2522 start to receive compilation warnings about `#![feature]` directives which enabled
2523 the new feature (because the directive is no longer necessary). However, if a
2524 feature is decided to be removed from the language, errors will be issued (if
2525 there isn't a parser error first). The directive in this case is no longer
2526 necessary, and it's likely that existing code will break if the feature isn't
2529 If an unknown feature is found in a directive, it results in a compiler error.
2530 An unknown feature is one which has never been recognized by the compiler.
2532 # Statements and expressions
2534 Rust is _primarily_ an expression language. This means that most forms of
2535 value-producing or effect-causing evaluation are directed by the uniform syntax
2536 category of _expressions_. Each kind of expression can typically _nest_ within
2537 each other kind of expression, and rules for evaluation of expressions involve
2538 specifying both the value produced by the expression and the order in which its
2539 sub-expressions are themselves evaluated.
2541 In contrast, statements in Rust serve _mostly_ to contain and explicitly
2542 sequence expression evaluation.
2546 A _statement_ is a component of a block, which is in turn a component of an
2547 outer [expression](#expressions) or [function](#functions).
2549 Rust has two kinds of statement: [declaration
2550 statements](#declaration-statements) and [expression
2551 statements](#expression-statements).
2553 ### Declaration statements
2555 A _declaration statement_ is one that introduces one or more *names* into the
2556 enclosing statement block. The declared names may denote new variables or new
2559 #### Item declarations
2561 An _item declaration statement_ has a syntactic form identical to an
2562 [item](#items) declaration within a module. Declaring an item — a
2563 function, enumeration, struct, type, static, trait, implementation or module
2564 — locally within a statement block is simply a way of restricting its
2565 scope to a narrow region containing all of its uses; it is otherwise identical
2566 in meaning to declaring the item outside the statement block.
2568 > **Note**: there is no implicit capture of the function's dynamic environment when
2569 > declaring a function-local item.
2571 #### `let` statements
2573 A _`let` statement_ introduces a new set of variables, given by a pattern. The
2574 pattern may be followed by a type annotation, and/or an initializer expression.
2575 When no type annotation is given, the compiler will infer the type, or signal
2576 an error if insufficient type information is available for definite inference.
2577 Any variables introduced by a variable declaration are visible from the point of
2578 declaration until the end of the enclosing block scope.
2580 ### Expression statements
2582 An _expression statement_ is one that evaluates an [expression](#expressions)
2583 and ignores its result. The type of an expression statement `e;` is always
2584 `()`, regardless of the type of `e`. As a rule, an expression statement's
2585 purpose is to trigger the effects of evaluating its expression.
2589 An expression may have two roles: it always produces a *value*, and it may have
2590 *effects* (otherwise known as "side effects"). An expression *evaluates to* a
2591 value, and has effects during *evaluation*. Many expressions contain
2592 sub-expressions (operands). The meaning of each kind of expression dictates
2595 * Whether or not to evaluate the sub-expressions when evaluating the expression
2596 * The order in which to evaluate the sub-expressions
2597 * How to combine the sub-expressions' values to obtain the value of the expression
2599 In this way, the structure of expressions dictates the structure of execution.
2600 Blocks are just another kind of expression, so blocks, statements, expressions,
2601 and blocks again can recursively nest inside each other to an arbitrary depth.
2603 #### Lvalues, rvalues and temporaries
2605 Expressions are divided into two main categories: _lvalues_ and _rvalues_.
2606 Likewise within each expression, sub-expressions may occur in _lvalue context_
2607 or _rvalue context_. The evaluation of an expression depends both on its own
2608 category and the context it occurs within.
2610 An lvalue is an expression that represents a memory location. These expressions
2611 are [paths](#path-expressions) (which refer to local variables, function and
2612 method arguments, or static variables), dereferences (`*expr`), [indexing
2613 expressions](#index-expressions) (`expr[expr]`), and [field
2614 references](#field-expressions) (`expr.f`). All other expressions are rvalues.
2616 The left operand of an [assignment](#assignment-expressions) or
2617 [compound-assignment](#compound-assignment-expressions) expression is
2618 an lvalue context, as is the single operand of a unary
2619 [borrow](#unary-operator-expressions). The discriminant or subject of
2620 a [match expression](#match-expressions) may be an lvalue context, if
2621 ref bindings are made, but is otherwise an rvalue context. All other
2622 expression contexts are rvalue contexts.
2624 When an lvalue is evaluated in an _lvalue context_, it denotes a memory
2625 location; when evaluated in an _rvalue context_, it denotes the value held _in_
2626 that memory location.
2628 ##### Temporary lifetimes
2630 When an rvalue is used in an lvalue context, a temporary un-named
2631 lvalue is created and used instead. The lifetime of temporary values
2632 is typically the innermost enclosing statement; the tail expression of
2633 a block is considered part of the statement that encloses the block.
2635 When a temporary rvalue is being created that is assigned into a `let`
2636 declaration, however, the temporary is created with the lifetime of
2637 the enclosing block instead, as using the enclosing statement (the
2638 `let` declaration) would be a guaranteed error (since a pointer to the
2639 temporary would be stored into a variable, but the temporary would be
2640 freed before the variable could be used). The compiler uses simple
2641 syntactic rules to decide which values are being assigned into a `let`
2642 binding, and therefore deserve a longer temporary lifetime.
2644 Here are some examples:
2646 - `let x = foo(&temp())`. The expression `temp()` is an rvalue. As it
2647 is being borrowed, a temporary is created which will be freed after
2648 the innermost enclosing statement (the `let` declaration, in this case).
2649 - `let x = temp().foo()`. This is the same as the previous example,
2650 except that the value of `temp()` is being borrowed via autoref on a
2651 method-call. Here we are assuming that `foo()` is an `&self` method
2652 defined in some trait, say `Foo`. In other words, the expression
2653 `temp().foo()` is equivalent to `Foo::foo(&temp())`.
2654 - `let x = &temp()`. Here, the same temporary is being assigned into
2655 `x`, rather than being passed as a parameter, and hence the
2656 temporary's lifetime is considered to be the enclosing block.
2657 - `let x = SomeStruct { foo: &temp() }`. As in the previous case, the
2658 temporary is assigned into a struct which is then assigned into a
2659 binding, and hence it is given the lifetime of the enclosing block.
2660 - `let x = [ &temp() ]`. As in the previous case, the
2661 temporary is assigned into an array which is then assigned into a
2662 binding, and hence it is given the lifetime of the enclosing block.
2663 - `let ref x = temp()`. In this case, the temporary is created using a ref binding,
2664 but the result is the same: the lifetime is extended to the enclosing block.
2666 #### Moved and copied types
2668 When a [local variable](#variables) is used as an
2669 [rvalue](#lvalues-rvalues-and-temporaries), the variable will be copied
2670 if its type implements `Copy`. All others are moved.
2672 ### Literal expressions
2674 A _literal expression_ consists of one of the [literal](#literals) forms
2675 described earlier. It directly describes a number, character, string, boolean
2676 value, or the unit value.
2680 "hello"; // string type
2681 '5'; // character type
2685 ### Path expressions
2687 A [path](#paths) used as an expression context denotes either a local variable
2688 or an item. Path expressions are [lvalues](#lvalues-rvalues-and-temporaries).
2690 ### Tuple expressions
2692 Tuples are written by enclosing zero or more comma-separated expressions in
2693 parentheses. They are used to create [tuple-typed](#tuple-types) values.
2697 ("a", 4usize, true);
2700 You can disambiguate a single-element tuple from a value in parentheses with a
2704 (0,); // single-element tuple
2705 (0); // zero in parentheses
2708 ### Struct expressions
2710 There are several forms of struct expressions. A _struct expression_
2711 consists of the [path](#paths) of a [struct item](#structs), followed by
2712 a brace-enclosed list of zero or more comma-separated name-value pairs,
2713 providing the field values of a new instance of the struct. A field name
2714 can be any identifier, and is separated from its value expression by a colon.
2715 The location denoted by a struct field is mutable if and only if the
2716 enclosing struct is mutable.
2718 A _tuple struct expression_ consists of the [path](#paths) of a [struct
2719 item](#structs), followed by a parenthesized list of one or more
2720 comma-separated expressions (in other words, the path of a struct item
2721 followed by a tuple expression). The struct item must be a tuple struct
2724 A _unit-like struct expression_ consists only of the [path](#paths) of a
2725 [struct item](#structs).
2727 The following are examples of struct expressions:
2730 # struct Point { x: f64, y: f64 }
2731 # struct NothingInMe { }
2732 # struct TuplePoint(f64, f64);
2733 # mod game { pub struct User<'a> { pub name: &'a str, pub age: u32, pub score: usize } }
2734 # struct Cookie; fn some_fn<T>(t: T) {}
2735 Point {x: 10.0, y: 20.0};
2737 TuplePoint(10.0, 20.0);
2738 let u = game::User {name: "Joe", age: 35, score: 100_000};
2739 some_fn::<Cookie>(Cookie);
2742 A struct expression forms a new value of the named struct type. Note
2743 that for a given *unit-like* struct type, this will always be the same
2746 A struct expression can terminate with the syntax `..` followed by an
2747 expression to denote a functional update. The expression following `..` (the
2748 base) must have the same struct type as the new struct type being formed.
2749 The entire expression denotes the result of constructing a new struct (with
2750 the same type as the base expression) with the given values for the fields that
2751 were explicitly specified and the values in the base expression for all other
2755 # struct Point3d { x: i32, y: i32, z: i32 }
2756 let base = Point3d {x: 1, y: 2, z: 3};
2757 Point3d {y: 0, z: 10, .. base};
2760 ### Block expressions
2762 A _block expression_ is similar to a module in terms of the declarations that
2763 are possible. Each block conceptually introduces a new namespace scope. Use
2764 items can bring new names into scopes and declared items are in scope for only
2767 A block will execute each statement sequentially, and then execute the
2768 expression (if given). If the block ends in a statement, its value is `()`:
2771 let x: () = { println!("Hello."); };
2774 If it ends in an expression, its value and type are that of the expression:
2777 let x: i32 = { println!("Hello."); 5 };
2782 ### Method-call expressions
2784 A _method call_ consists of an expression followed by a single dot, an
2785 identifier, and a parenthesized expression-list. Method calls are resolved to
2786 methods on specific traits, either statically dispatching to a method if the
2787 exact `self`-type of the left-hand-side is known, or dynamically dispatching if
2788 the left-hand-side expression is an indirect [trait object](#trait-objects).
2790 ### Field expressions
2792 A _field expression_ consists of an expression followed by a single dot and an
2793 identifier, when not immediately followed by a parenthesized expression-list
2794 (the latter is a [method call expression](#method-call-expressions)). A field
2795 expression denotes a field of a [struct](#struct-types).
2800 (Struct {a: 10, b: 20}).a;
2803 A field access is an [lvalue](#lvalues-rvalues-and-temporaries) referring to
2804 the value of that field. When the type providing the field inherits mutability,
2805 it can be [assigned](#assignment-expressions) to.
2807 Also, if the type of the expression to the left of the dot is a
2808 pointer, it is automatically dereferenced as many times as necessary
2809 to make the field access possible. In cases of ambiguity, we prefer
2810 fewer autoderefs to more.
2812 ### Array expressions
2814 An [array](#array-and-slice-types) _expression_ is written by enclosing zero
2815 or more comma-separated expressions of uniform type in square brackets.
2817 In the `[expr ';' expr]` form, the expression after the `';'` must be a
2818 constant expression that can be evaluated at compile time, such as a
2819 [literal](#literals) or a [static item](#static-items).
2823 ["a", "b", "c", "d"];
2824 [0; 128]; // array with 128 zeros
2825 [0u8, 0u8, 0u8, 0u8];
2828 ### Index expressions
2830 [Array](#array-and-slice-types)-typed expressions can be indexed by
2831 writing a square-bracket-enclosed expression (the index) after them. When the
2832 array is mutable, the resulting [lvalue](#lvalues-rvalues-and-temporaries) can
2835 Indices are zero-based, and may be of any integral type. Vector access is
2836 bounds-checked at compile-time for constant arrays being accessed with a constant index value.
2837 Otherwise a check will be performed at run-time that will put the thread in a _panicked state_ if it fails.
2842 let x = (["a", "b"])[10]; // compiler error: const index-expr is out of bounds
2845 let y = (["a", "b"])[n]; // panics
2847 let arr = ["a", "b"];
2851 Also, if the type of the expression to the left of the brackets is a
2852 pointer, it is automatically dereferenced as many times as necessary
2853 to make the indexing possible. In cases of ambiguity, we prefer fewer
2856 ### Range expressions
2858 The `..` operator will construct an object of one of the `std::ops::Range` variants.
2861 1..2; // std::ops::Range
2862 3..; // std::ops::RangeFrom
2863 ..4; // std::ops::RangeTo
2864 ..; // std::ops::RangeFull
2867 The following expressions are equivalent.
2870 let x = std::ops::Range {start: 0, end: 10};
2876 Similarly, the `...` operator will construct an object of one of the
2877 `std::ops::RangeInclusive` variants.
2880 # #![feature(inclusive_range_syntax)]
2881 1...2; // std::ops::RangeInclusive
2882 ...4; // std::ops::RangeToInclusive
2885 The following expressions are equivalent.
2888 # #![feature(inclusive_range_syntax, inclusive_range)]
2889 let x = std::ops::RangeInclusive::NonEmpty {start: 0, end: 10};
2895 ### Unary operator expressions
2897 Rust defines the following unary operators. With the exception of `?`, they are
2898 all written as prefix operators, before the expression they apply to.
2901 : Negation. Signed integer types and floating-point types support negation. It
2902 is an error to apply negation to unsigned types; for example, the compiler
2905 : Dereference. When applied to a [pointer](#pointer-types) it denotes the
2906 pointed-to location. For pointers to mutable locations, the resulting
2907 [lvalue](#lvalues-rvalues-and-temporaries) can be assigned to.
2908 On non-pointer types, it calls the `deref` method of the `std::ops::Deref`
2909 trait, or the `deref_mut` method of the `std::ops::DerefMut` trait (if
2910 implemented by the type and required for an outer expression that will or
2911 could mutate the dereference), and produces the result of dereferencing the
2912 `&` or `&mut` borrowed pointer returned from the overload method.
2914 : Logical negation. On the boolean type, this flips between `true` and
2915 `false`. On integer types, this inverts the individual bits in the
2916 two's complement representation of the value.
2918 : Borrowing. When applied to an lvalue, these operators produce a
2919 reference (pointer) to the lvalue. The lvalue is also placed into
2920 a borrowed state for the duration of the reference. For a shared
2921 borrow (`&`), this implies that the lvalue may not be mutated, but
2922 it may be read or shared again. For a mutable borrow (`&mut`), the
2923 lvalue may not be accessed in any way until the borrow expires.
2924 If the `&` or `&mut` operators are applied to an rvalue, a
2925 temporary value is created; the lifetime of this temporary value
2926 is defined by [syntactic rules](#temporary-lifetimes).
2928 : Propagating errors if applied to `Err(_)` and unwrapping if
2929 applied to `Ok(_)`. Only works on the `Result<T, E>` type,
2930 and written in postfix notation.
2932 ### Binary operator expressions
2934 Binary operators expressions are given in terms of [operator
2935 precedence](#operator-precedence).
2937 #### Arithmetic operators
2939 Binary arithmetic expressions are syntactic sugar for calls to built-in traits,
2940 defined in the `std::ops` module of the `std` library. This means that
2941 arithmetic operators can be overridden for user-defined types. The default
2942 meaning of the operators on standard types is given here.
2945 : Addition and array/string concatenation.
2946 Calls the `add` method on the `std::ops::Add` trait.
2949 Calls the `sub` method on the `std::ops::Sub` trait.
2952 Calls the `mul` method on the `std::ops::Mul` trait.
2955 Calls the `div` method on the `std::ops::Div` trait.
2958 Calls the `rem` method on the `std::ops::Rem` trait.
2960 #### Bitwise operators
2962 Like the [arithmetic operators](#arithmetic-operators), bitwise operators are
2963 syntactic sugar for calls to methods of built-in traits. This means that
2964 bitwise operators can be overridden for user-defined types. The default
2965 meaning of the operators on standard types is given here. Bitwise `&`, `|` and
2966 `^` applied to boolean arguments are equivalent to logical `&&`, `||` and `!=`
2967 evaluated in non-lazy fashion.
2971 Calls the `bitand` method of the `std::ops::BitAnd` trait.
2973 : Bitwise inclusive OR.
2974 Calls the `bitor` method of the `std::ops::BitOr` trait.
2976 : Bitwise exclusive OR.
2977 Calls the `bitxor` method of the `std::ops::BitXor` trait.
2980 Calls the `shl` method of the `std::ops::Shl` trait.
2982 : Right shift (arithmetic).
2983 Calls the `shr` method of the `std::ops::Shr` trait.
2985 #### Lazy boolean operators
2987 The operators `||` and `&&` may be applied to operands of boolean type. The
2988 `||` operator denotes logical 'or', and the `&&` operator denotes logical
2989 'and'. They differ from `|` and `&` in that the right-hand operand is only
2990 evaluated when the left-hand operand does not already determine the result of
2991 the expression. That is, `||` only evaluates its right-hand operand when the
2992 left-hand operand evaluates to `false`, and `&&` only when it evaluates to
2995 #### Comparison operators
2997 Comparison operators are, like the [arithmetic
2998 operators](#arithmetic-operators), and [bitwise operators](#bitwise-operators),
2999 syntactic sugar for calls to built-in traits. This means that comparison
3000 operators can be overridden for user-defined types. The default meaning of the
3001 operators on standard types is given here.
3005 Calls the `eq` method on the `std::cmp::PartialEq` trait.
3008 Calls the `ne` method on the `std::cmp::PartialEq` trait.
3011 Calls the `lt` method on the `std::cmp::PartialOrd` trait.
3014 Calls the `gt` method on the `std::cmp::PartialOrd` trait.
3016 : Less than or equal.
3017 Calls the `le` method on the `std::cmp::PartialOrd` trait.
3019 : Greater than or equal.
3020 Calls the `ge` method on the `std::cmp::PartialOrd` trait.
3022 #### Type cast expressions
3024 A type cast expression is denoted with the binary operator `as`.
3026 Executing an `as` expression casts the value on the left-hand side to the type
3027 on the right-hand side.
3029 An example of an `as` expression:
3032 # fn sum(values: &[f64]) -> f64 { 0.0 }
3033 # fn len(values: &[f64]) -> i32 { 0 }
3035 fn average(values: &[f64]) -> f64 {
3036 let sum: f64 = sum(values);
3037 let size: f64 = len(values) as f64;
3042 Some of the conversions which can be done through the `as` operator
3043 can also be done implicitly at various points in the program, such as
3044 argument passing and assignment to a `let` binding with an explicit
3045 type. Implicit conversions are limited to "harmless" conversions that
3046 do not lose information and which have minimal or no risk of
3047 surprising side-effects on the dynamic execution semantics.
3049 #### Assignment expressions
3051 An _assignment expression_ consists of an
3052 [lvalue](#lvalues-rvalues-and-temporaries) expression followed by an equals
3053 sign (`=`) and an [rvalue](#lvalues-rvalues-and-temporaries) expression.
3055 Evaluating an assignment expression [either copies or
3056 moves](#moved-and-copied-types) its right-hand operand to its left-hand
3065 #### Compound assignment expressions
3067 The `+`, `-`, `*`, `/`, `%`, `&`, `|`, `^`, `<<`, and `>>` operators may be
3068 composed with the `=` operator. The expression `lval OP= val` is equivalent to
3069 `lval = lval OP val`. For example, `x = x + 1` may be written as `x += 1`.
3071 Any such expression always has the [`unit`](#tuple-types) type.
3073 #### Operator precedence
3075 The precedence of Rust binary operators is ordered as follows, going from
3078 ```{.text .precedence}
3094 Operators at the same precedence level are evaluated left-to-right. [Unary
3095 operators](#unary-operator-expressions) have the same precedence level and are
3096 stronger than any of the binary operators.
3098 ### Grouped expressions
3100 An expression enclosed in parentheses evaluates to the result of the enclosed
3101 expression. Parentheses can be used to explicitly specify evaluation order
3102 within an expression.
3104 An example of a parenthesized expression:
3107 let x: i32 = (2 + 3) * 4;
3111 ### Call expressions
3113 A _call expression_ invokes a function, providing zero or more input variables
3114 and an optional location to move the function's output into. If the function
3115 eventually returns, then the expression completes.
3117 Some examples of call expressions:
3120 # fn add(x: i32, y: i32) -> i32 { 0 }
3122 let x: i32 = add(1i32, 2i32);
3123 let pi: Result<f32, _> = "3.14".parse();
3126 ### Lambda expressions
3128 A _lambda expression_ (sometimes called an "anonymous function expression")
3129 defines a function and denotes it as a value, in a single expression. A lambda
3130 expression is a pipe-symbol-delimited (`|`) list of identifiers followed by an
3133 A lambda expression denotes a function that maps a list of parameters
3134 (`ident_list`) onto the expression that follows the `ident_list`. The
3135 identifiers in the `ident_list` are the parameters to the function. These
3136 parameters' types need not be specified, as the compiler infers them from
3139 Lambda expressions are most useful when passing functions as arguments to other
3140 functions, as an abbreviation for defining and capturing a separate function.
3142 Significantly, lambda expressions _capture their environment_, which regular
3143 [function definitions](#functions) do not. The exact type of capture depends
3144 on the [function type](#function-types) inferred for the lambda expression. In
3145 the simplest and least-expensive form (analogous to a ```|| { }``` expression),
3146 the lambda expression captures its environment by reference, effectively
3147 borrowing pointers to all outer variables mentioned inside the function.
3148 Alternately, the compiler may infer that a lambda expression should copy or
3149 move values (depending on their type) from the environment into the lambda
3150 expression's captured environment. A lambda can be forced to capture its
3151 environment by moving values by prefixing it with the `move` keyword.
3153 In this example, we define a function `ten_times` that takes a higher-order
3154 function argument, and we then call it with a lambda expression as an argument,
3155 followed by a lambda expression that moves values from its environment.
3158 fn ten_times<F>(f: F) where F: Fn(i32) {
3159 for index in 0..10 {
3164 ten_times(|j| println!("hello, {}", j));
3166 let word = "konnichiwa".to_owned();
3167 ten_times(move |j| println!("{}, {}", word, j));
3172 A `loop` expression denotes an infinite loop.
3174 A `loop` expression may optionally have a _label_. The label is written as
3175 a lifetime preceding the loop expression, as in `'foo: loop{ }`. If a
3176 label is present, then labeled `break` and `continue` expressions nested
3177 within this loop may exit out of this loop or return control to its head.
3178 See [break expressions](#break-expressions) and [continue
3179 expressions](#continue-expressions).
3181 ### `break` expressions
3183 A `break` expression has an optional _label_. If the label is absent, then
3184 executing a `break` expression immediately terminates the innermost loop
3185 enclosing it. It is only permitted in the body of a loop. If the label is
3186 present, then `break 'foo` terminates the loop with label `'foo`, which need not
3187 be the innermost label enclosing the `break` expression, but must enclose it.
3189 ### `continue` expressions
3191 A `continue` expression has an optional _label_. If the label is absent, then
3192 executing a `continue` expression immediately terminates the current iteration
3193 of the innermost loop enclosing it, returning control to the loop *head*. In
3194 the case of a `while` loop, the head is the conditional expression controlling
3195 the loop. In the case of a `for` loop, the head is the call-expression
3196 controlling the loop. If the label is present, then `continue 'foo` returns
3197 control to the head of the loop with label `'foo`, which need not be the
3198 innermost label enclosing the `continue` expression, but must enclose it.
3200 A `continue` expression is only permitted in the body of a loop.
3204 A `while` loop begins by evaluating the boolean loop conditional expression.
3205 If the loop conditional expression evaluates to `true`, the loop body block
3206 executes and control returns to the loop conditional expression. If the loop
3207 conditional expression evaluates to `false`, the `while` expression completes.
3220 Like `loop` expressions, `while` loops can be controlled with `break` or
3221 `continue`, and may optionally have a _label_. See [infinite
3222 loops](#infinite-loops), [break expressions](#break-expressions), and
3223 [continue expressions](#continue-expressions) for more information.
3225 ### `for` expressions
3227 A `for` expression is a syntactic construct for looping over elements provided
3228 by an implementation of `std::iter::IntoIterator`.
3230 An example of a `for` loop over the contents of an array:
3234 # fn bar(f: &Foo) { }
3239 let v: &[Foo] = &[a, b, c];
3246 An example of a for loop over a series of integers:
3249 # fn bar(b:usize) { }
3255 Like `loop` expressions, `for` loops can be controlled with `break` or
3256 `continue`, and may optionally have a _label_. See [infinite
3257 loops](#infinite-loops), [break expressions](#break-expressions), and
3258 [continue expressions](#continue-expressions) for more information.
3260 ### `if` expressions
3262 An `if` expression is a conditional branch in program control. The form of an
3263 `if` expression is a condition expression, followed by a consequent block, any
3264 number of `else if` conditions and blocks, and an optional trailing `else`
3265 block. The condition expressions must have type `bool`. If a condition
3266 expression evaluates to `true`, the consequent block is executed and any
3267 subsequent `else if` or `else` block is skipped. If a condition expression
3268 evaluates to `false`, the consequent block is skipped and any subsequent `else
3269 if` condition is evaluated. If all `if` and `else if` conditions evaluate to
3270 `false` then any `else` block is executed.
3272 ### `match` expressions
3274 A `match` expression branches on a *pattern*. The exact form of matching that
3275 occurs depends on the pattern. Patterns consist of some combination of
3276 literals, destructured arrays or enum constructors, structs and tuples,
3277 variable binding specifications, wildcards (`..`), and placeholders (`_`). A
3278 `match` expression has a *head expression*, which is the value to compare to
3279 the patterns. The type of the patterns must equal the type of the head
3282 In a pattern whose head expression has an `enum` type, a placeholder (`_`)
3283 stands for a *single* data field, whereas a wildcard `..` stands for *all* the
3284 fields of a particular variant.
3286 A `match` behaves differently depending on whether or not the head expression
3287 is an [lvalue or an rvalue](#lvalues-rvalues-and-temporaries). If the head
3288 expression is an rvalue, it is first evaluated into a temporary location, and
3289 the resulting value is sequentially compared to the patterns in the arms until
3290 a match is found. The first arm with a matching pattern is chosen as the branch
3291 target of the `match`, any variables bound by the pattern are assigned to local
3292 variables in the arm's block, and control enters the block.
3294 When the head expression is an lvalue, the match does not allocate a temporary
3295 location (however, a by-value binding may copy or move from the lvalue). When
3296 possible, it is preferable to match on lvalues, as the lifetime of these
3297 matches inherits the lifetime of the lvalue, rather than being restricted to
3298 the inside of the match.
3300 An example of a `match` expression:
3306 1 => println!("one"),
3307 2 => println!("two"),
3308 3 => println!("three"),
3309 4 => println!("four"),
3310 5 => println!("five"),
3311 _ => println!("something else"),
3315 Patterns that bind variables default to binding to a copy or move of the
3316 matched value (depending on the matched value's type). This can be changed to
3317 bind to a reference by using the `ref` keyword, or to a mutable reference using
3320 Subpatterns can also be bound to variables by the use of the syntax `variable @
3321 subpattern`. For example:
3327 e @ 1 ... 5 => println!("got a range element {}", e),
3328 _ => println!("anything"),
3332 Patterns can also dereference pointers by using the `&`, `&mut` and `box`
3333 symbols, as appropriate. For example, these two matches on `x: &i32` are
3338 let y = match *x { 0 => "zero", _ => "some" };
3339 let z = match x { &0 => "zero", _ => "some" };
3344 Multiple match patterns may be joined with the `|` operator. A range of values
3345 may be specified with `...`. For example:
3350 let message = match x {
3351 0 | 1 => "not many",
3357 Range patterns only work on scalar types (like integers and characters; not
3358 like arrays and structs, which have sub-components). A range pattern may not
3359 be a sub-range of another range pattern inside the same `match`.
3361 Finally, match patterns can accept *pattern guards* to further refine the
3362 criteria for matching a case. Pattern guards appear after the pattern and
3363 consist of a bool-typed expression following the `if` keyword. A pattern guard
3364 may refer to the variables bound within the pattern they follow.
3367 # let maybe_digit = Some(0);
3368 # fn process_digit(i: i32) { }
3369 # fn process_other(i: i32) { }
3371 let message = match maybe_digit {
3372 Some(x) if x < 10 => process_digit(x),
3373 Some(x) => process_other(x),
3378 ### `if let` expressions
3380 An `if let` expression is semantically identical to an `if` expression but in
3381 place of a condition expression it expects a `let` statement with a refutable
3382 pattern. If the value of the expression on the right hand side of the `let`
3383 statement matches the pattern, the corresponding block will execute, otherwise
3384 flow proceeds to the first `else` block that follows.
3387 let dish = ("Ham", "Eggs");
3389 // this body will be skipped because the pattern is refuted
3390 if let ("Bacon", b) = dish {
3391 println!("Bacon is served with {}", b);
3394 // this body will execute
3395 if let ("Ham", b) = dish {
3396 println!("Ham is served with {}", b);
3400 ### `while let` loops
3402 A `while let` loop is semantically identical to a `while` loop but in place of
3403 a condition expression it expects `let` statement with a refutable pattern. If
3404 the value of the expression on the right hand side of the `let` statement
3405 matches the pattern, the loop body block executes and control returns to the
3406 pattern matching statement. Otherwise, the while expression completes.
3408 ### `return` expressions
3410 Return expressions are denoted with the keyword `return`. Evaluating a `return`
3411 expression moves its argument into the designated output location for the
3412 current function call, destroys the current function activation frame, and
3413 transfers control to the caller frame.
3415 An example of a `return` expression:
3418 fn max(a: i32, b: i32) -> i32 {
3430 Every variable, item and value in a Rust program has a type. The _type_ of a
3431 *value* defines the interpretation of the memory holding it.
3433 Built-in types and type-constructors are tightly integrated into the language,
3434 in nontrivial ways that are not possible to emulate in user-defined types.
3435 User-defined types have limited capabilities.
3439 The primitive types are the following:
3441 * The boolean type `bool` with values `true` and `false`.
3442 * The machine types (integer and floating-point).
3443 * The machine-dependent integer types.
3451 The machine types are the following:
3453 * The unsigned word types `u8`, `u16`, `u32` and `u64`, with values drawn from
3454 the integer intervals [0, 2^8 - 1], [0, 2^16 - 1], [0, 2^32 - 1] and
3455 [0, 2^64 - 1] respectively.
3457 * The signed two's complement word types `i8`, `i16`, `i32` and `i64`, with
3458 values drawn from the integer intervals [-(2^(7)), 2^7 - 1],
3459 [-(2^(15)), 2^15 - 1], [-(2^(31)), 2^31 - 1], [-(2^(63)), 2^63 - 1]
3462 * The IEEE 754-2008 `binary32` and `binary64` floating-point types: `f32` and
3463 `f64`, respectively.
3465 #### Machine-dependent integer types
3467 The `usize` type is an unsigned integer type with the same number of bits as the
3468 platform's pointer type. It can represent every memory address in the process.
3470 The `isize` type is a signed integer type with the same number of bits as the
3471 platform's pointer type. The theoretical upper bound on object and array size
3472 is the maximum `isize` value. This ensures that `isize` can be used to calculate
3473 differences between pointers into an object or array and can address every byte
3474 within an object along with one byte past the end.
3478 The types `char` and `str` hold textual data.
3480 A value of type `char` is a [Unicode scalar value](
3481 http://www.unicode.org/glossary/#unicode_scalar_value) (i.e. a code point that
3482 is not a surrogate), represented as a 32-bit unsigned word in the 0x0000 to
3483 0xD7FF or 0xE000 to 0x10FFFF range. A `[char]` array is effectively an UCS-4 /
3486 A value of type `str` is a Unicode string, represented as an array of 8-bit
3487 unsigned bytes holding a sequence of UTF-8 code points. Since `str` is of
3488 unknown size, it is not a _first-class_ type, but can only be instantiated
3489 through a pointer type, such as `&str`.
3493 A tuple *type* is a heterogeneous product of other types, called the *elements*
3494 of the tuple. It has no nominal name and is instead structurally typed.
3496 Tuple types and values are denoted by listing the types or values of their
3497 elements, respectively, in a parenthesized, comma-separated list.
3499 Because tuple elements don't have a name, they can only be accessed by
3500 pattern-matching or by using `N` directly as a field to access the
3503 An example of a tuple type and its use:
3506 type Pair<'a> = (i32, &'a str);
3507 let p: Pair<'static> = (10, "ten");
3511 assert_eq!(b, "ten");
3512 assert_eq!(p.0, 10);
3513 assert_eq!(p.1, "ten");
3516 For historical reasons and convenience, the tuple type with no elements (`()`)
3517 is often called ‘unit’ or ‘the unit type’.
3519 ### Array, and Slice types
3521 Rust has two different types for a list of items:
3523 * `[T; N]`, an 'array'
3526 An array has a fixed size, and can be allocated on either the stack or the
3529 A slice is a 'view' into an array. It doesn't own the data it points
3535 // A stack-allocated array
3536 let array: [i32; 3] = [1, 2, 3];
3538 // A heap-allocated array
3539 let vector: Vec<i32> = vec![1, 2, 3];
3541 // A slice into an array
3542 let slice: &[i32] = &vector[..];
3545 As you can see, the `vec!` macro allows you to create a `Vec<T>` easily. The
3546 `vec!` macro is also part of the standard library, rather than the language.
3548 All in-bounds elements of arrays and slices are always initialized, and access
3549 to an array or slice is always bounds-checked.
3553 A `struct` *type* is a heterogeneous product of other types, called the
3554 *fields* of the type.[^structtype]
3556 [^structtype]: `struct` types are analogous to `struct` types in C,
3557 the *record* types of the ML family,
3558 or the *struct* types of the Lisp family.
3560 New instances of a `struct` can be constructed with a [struct
3561 expression](#struct-expressions).
3563 The memory layout of a `struct` is undefined by default to allow for compiler
3564 optimizations like field reordering, but it can be fixed with the
3565 `#[repr(...)]` attribute. In either case, fields may be given in any order in
3566 a corresponding struct *expression*; the resulting `struct` value will always
3567 have the same memory layout.
3569 The fields of a `struct` may be qualified by [visibility
3570 modifiers](#visibility-and-privacy), to allow access to data in a
3571 struct outside a module.
3573 A _tuple struct_ type is just like a struct type, except that the fields are
3576 A _unit-like struct_ type is like a struct type, except that it has no
3577 fields. The one value constructed by the associated [struct
3578 expression](#struct-expressions) is the only value that inhabits such a
3581 ### Enumerated types
3583 An *enumerated type* is a nominal, heterogeneous disjoint union type, denoted
3584 by the name of an [`enum` item](#enumerations). [^enumtype]
3586 [^enumtype]: The `enum` type is analogous to a `data` constructor declaration in
3587 ML, or a *pick ADT* in Limbo.
3589 An [`enum` item](#enumerations) declares both the type and a number of *variant
3590 constructors*, each of which is independently named and takes an optional tuple
3593 New instances of an `enum` can be constructed by calling one of the variant
3594 constructors, in a [call expression](#call-expressions).
3596 Any `enum` value consumes as much memory as the largest variant constructor for
3597 its corresponding `enum` type.
3599 Enum types cannot be denoted *structurally* as types, but must be denoted by
3600 named reference to an [`enum` item](#enumerations).
3604 Nominal types — [enumerations](#enumerated-types) and
3605 [structs](#struct-types) — may be recursive. That is, each `enum`
3606 constructor or `struct` field may refer, directly or indirectly, to the
3607 enclosing `enum` or `struct` type itself. Such recursion has restrictions:
3609 * Recursive types must include a nominal type in the recursion
3610 (not mere [type definitions](grammar.html#type-definitions),
3611 or other structural types such as [arrays](#array-and-slice-types) or [tuples](#tuple-types)).
3612 * A recursive `enum` item must have at least one non-recursive constructor
3613 (in order to give the recursion a basis case).
3614 * The size of a recursive type must be finite;
3615 in other words the recursive fields of the type must be [pointer types](#pointer-types).
3616 * Recursive type definitions can cross module boundaries, but not module *visibility* boundaries,
3617 or crate boundaries (in order to simplify the module system and type checker).
3619 An example of a *recursive* type and its use:
3624 Cons(T, Box<List<T>>)
3627 let a: List<i32> = List::Cons(7, Box::new(List::Cons(13, Box::new(List::Nil))));
3632 All pointers in Rust are explicit first-class values. They can be copied,
3633 stored into data structs, and returned from functions. There are two
3634 varieties of pointer in Rust:
3637 : These point to memory _owned by some other value_.
3638 A reference type is written `&type`,
3639 or `&'a type` when you need to specify an explicit lifetime.
3640 Copying a reference is a "shallow" operation:
3641 it involves only copying the pointer itself.
3642 Releasing a reference has no effect on the value it points to,
3643 but a reference of a temporary value will keep it alive during the scope
3644 of the reference itself.
3646 * Raw pointers (`*`)
3647 : Raw pointers are pointers without safety or liveness guarantees.
3648 Raw pointers are written as `*const T` or `*mut T`,
3649 for example `*const i32` means a raw pointer to a 32-bit integer.
3650 Copying or dropping a raw pointer has no effect on the lifecycle of any
3651 other value. Dereferencing a raw pointer or converting it to any other
3652 pointer type is an [`unsafe` operation](#unsafe-functions).
3653 Raw pointers are generally discouraged in Rust code;
3654 they exist to support interoperability with foreign code,
3655 and writing performance-critical or low-level functions.
3657 The standard library contains additional 'smart pointer' types beyond references
3662 The function type constructor `fn` forms new function types. A function type
3663 consists of a possibly-empty set of function-type modifiers (such as `unsafe`
3664 or `extern`), a sequence of input types and an output type.
3666 An example of a `fn` type:
3669 fn add(x: i32, y: i32) -> i32 {
3673 let mut x = add(5,7);
3675 type Binop = fn(i32, i32) -> i32;
3676 let bo: Binop = add;
3680 #### Function types for specific items
3682 Internal to the compiler, there are also function types that are specific to a particular
3683 function item. In the following snippet, for example, the internal types of the functions
3684 `foo` and `bar` are different, despite the fact that they have the same signature:
3691 The types of `foo` and `bar` can both be implicitly coerced to the fn
3692 pointer type `fn()`. There is currently no syntax for unique fn types,
3693 though the compiler will emit a type like `fn() {foo}` in error
3694 messages to indicate "the unique fn type for the function `foo`".
3698 A [lambda expression](#lambda-expressions) produces a closure value with
3699 a unique, anonymous type that cannot be written out.
3701 Depending on the requirements of the closure, its type implements one or
3702 more of the closure traits:
3705 : The closure can be called once. A closure called as `FnOnce`
3706 can move out values from its environment.
3709 : The closure can be called multiple times as mutable. A closure called as
3710 `FnMut` can mutate values from its environment. `FnMut` inherits from
3711 `FnOnce` (i.e. anything implementing `FnMut` also implements `FnOnce`).
3714 : The closure can be called multiple times through a shared reference.
3715 A closure called as `Fn` can neither move out from nor mutate values
3716 from its environment. `Fn` inherits from `FnMut`, which itself
3717 inherits from `FnOnce`.
3722 In Rust, a type like `&SomeTrait` or `Box<SomeTrait>` is called a _trait object_.
3723 Each instance of a trait object includes:
3725 - a pointer to an instance of a type `T` that implements `SomeTrait`
3726 - a _virtual method table_, often just called a _vtable_, which contains, for
3727 each method of `SomeTrait` that `T` implements, a pointer to `T`'s
3728 implementation (i.e. a function pointer).
3730 The purpose of trait objects is to permit "late binding" of methods. Calling a
3731 method on a trait object results in virtual dispatch at runtime: that is, a
3732 function pointer is loaded from the trait object vtable and invoked indirectly.
3733 The actual implementation for each vtable entry can vary on an object-by-object
3736 Note that for a trait object to be instantiated, the trait must be
3737 _object-safe_. Object safety rules are defined in [RFC 255].
3739 [RFC 255]: https://github.com/rust-lang/rfcs/blob/master/text/0255-object-safety.md
3741 Given a pointer-typed expression `E` of type `&T` or `Box<T>`, where `T`
3742 implements trait `R`, casting `E` to the corresponding pointer type `&R` or
3743 `Box<R>` results in a value of the _trait object_ `R`. This result is
3744 represented as a pair of pointers: the vtable pointer for the `T`
3745 implementation of `R`, and the pointer value of `E`.
3747 An example of a trait object:
3751 fn stringify(&self) -> String;
3754 impl Printable for i32 {
3755 fn stringify(&self) -> String { self.to_string() }
3758 fn print(a: Box<Printable>) {
3759 println!("{}", a.stringify());
3763 print(Box::new(10) as Box<Printable>);
3767 In this example, the trait `Printable` occurs as a trait object in both the
3768 type signature of `print`, and the cast expression in `main`.
3772 Within the body of an item that has type parameter declarations, the names of
3773 its type parameters are types:
3776 fn to_vec<A: Clone>(xs: &[A]) -> Vec<A> {
3780 let first: A = xs[0].clone();
3781 let mut rest: Vec<A> = to_vec(&xs[1..]);
3782 rest.insert(0, first);
3787 Here, `first` has type `A`, referring to `to_vec`'s `A` type parameter; and `rest`
3788 has type `Vec<A>`, a vector with element type `A`.
3792 The special type `Self` has a meaning within traits and impls. In a trait definition, it refers
3793 to an implicit type parameter representing the "implementing" type. In an impl,
3794 it is an alias for the implementing type. For example, in:
3801 impl From<i32> for String {
3802 fn from(x: i32) -> Self {
3808 The notation `Self` in the impl refers to the implementing type: `String`. In another
3813 fn make_string(&self) -> String;
3816 impl Printable for String {
3817 fn make_string(&self) -> String {
3823 The notation `&self` is a shorthand for `self: &Self`. In this case,
3824 in the impl, `Self` refers to the value of type `String` that is the
3825 receiver for a call to the method `make_string`.
3829 Subtyping is implicit and can occur at any stage in type checking or
3830 inference. Subtyping in Rust is very restricted and occurs only due to
3831 variance with respect to lifetimes and between types with higher ranked
3832 lifetimes. If we were to erase lifetimes from types, then the only subtyping
3833 would be due to type equality.
3835 Consider the following example: string literals always have `'static`
3836 lifetime. Nevertheless, we can assign `s` to `t`:
3840 let s: &'static str = "hi";
3844 Since `'static` "lives longer" than `'a`, `&'static str` is a subtype of
3849 Coercions are defined in [RFC 401]. A coercion is implicit and has no syntax.
3851 [RFC 401]: https://github.com/rust-lang/rfcs/blob/master/text/0401-coercions.md
3855 A coercion can only occur at certain coercion sites in a program; these are
3856 typically places where the desired type is explicit or can be derived by
3857 propagation from explicit types (without type inference). Possible coercion
3860 * `let` statements where an explicit type is given.
3862 For example, `42` is coerced to have type `i8` in the following:
3868 * `static` and `const` statements (similar to `let` statements).
3870 * Arguments for function calls
3872 The value being coerced is the actual parameter, and it is coerced to
3873 the type of the formal parameter.
3875 For example, `42` is coerced to have type `i8` in the following:
3885 * Instantiations of struct or variant fields
3887 For example, `42` is coerced to have type `i8` in the following:
3890 struct Foo { x: i8 }
3897 * Function results, either the final line of a block if it is not
3898 semicolon-terminated or any expression in a `return` statement
3900 For example, `42` is coerced to have type `i8` in the following:
3908 If the expression in one of these coercion sites is a coercion-propagating
3909 expression, then the relevant sub-expressions in that expression are also
3910 coercion sites. Propagation recurses from these new coercion sites.
3911 Propagating expressions and their relevant sub-expressions are:
3913 * Array literals, where the array has type `[U; n]`. Each sub-expression in
3914 the array literal is a coercion site for coercion to type `U`.
3916 * Array literals with repeating syntax, where the array has type `[U; n]`. The
3917 repeated sub-expression is a coercion site for coercion to type `U`.
3919 * Tuples, where a tuple is a coercion site to type `(U_0, U_1, ..., U_n)`.
3920 Each sub-expression is a coercion site to the respective type, e.g. the
3921 zeroth sub-expression is a coercion site to type `U_0`.
3923 * Parenthesized sub-expressions (`(e)`): if the expression has type `U`, then
3924 the sub-expression is a coercion site to `U`.
3926 * Blocks: if a block has type `U`, then the last expression in the block (if
3927 it is not semicolon-terminated) is a coercion site to `U`. This includes
3928 blocks which are part of control flow statements, such as `if`/`else`, if
3929 the block has a known type.
3933 Coercion is allowed between the following types:
3935 * `T` to `U` if `T` is a subtype of `U` (*reflexive case*)
3937 * `T_1` to `T_3` where `T_1` coerces to `T_2` and `T_2` coerces to `T_3`
3940 Note that this is not fully supported yet
3944 * `*mut T` to `*const T`
3946 * `&T` to `*const T`
3948 * `&mut T` to `*mut T`
3950 * `&T` to `&U` if `T` implements `Deref<Target = U>`. For example:
3953 use std::ops::Deref;
3955 struct CharContainer {
3959 impl Deref for CharContainer {
3962 fn deref<'a>(&'a self) -> &'a char {
3967 fn foo(arg: &char) {}
3970 let x = &mut CharContainer { value: 'y' };
3971 foo(x); //&mut CharContainer is coerced to &char.
3975 * `&mut T` to `&mut U` if `T` implements `DerefMut<Target = U>`.
3977 * TyCtor(`T`) to TyCtor(coerce_inner(`T`)), where TyCtor(`T`) is one of
3985 - coerce_inner(`[T, ..n]`) = `[T]`
3986 - coerce_inner(`T`) = `U` where `T` is a concrete type which implements the
3989 In the future, coerce_inner will be recursively extended to tuples and
3990 structs. In addition, coercions from sub-traits to super-traits will be
3991 added. See [RFC 401] for more details.
3995 Several traits define special evaluation behavior.
3999 The `Copy` trait changes the semantics of a type implementing it. Values whose
4000 type implements `Copy` are copied rather than moved upon assignment.
4002 ## The `Sized` trait
4004 The `Sized` trait indicates that the size of this type is known at compile-time.
4008 The `Drop` trait provides a destructor, to be run whenever a value of this type
4011 ## The `Deref` trait
4013 The `Deref<Target = U>` trait allows a type to implicitly implement all the methods
4014 of the type `U`. When attempting to resolve a method call, the compiler will search
4015 the top-level type for the implementation of the called method. If no such method is
4016 found, `.deref()` is called and the compiler continues to search for the method
4017 implementation in the returned type `U`.
4021 The `Send` trait indicates that a value of this type is safe to send from one
4026 The `Sync` trait indicates that a value of this type is safe to share between
4031 A Rust program's memory consists of a static set of *items* and a *heap*.
4032 Immutable portions of the heap may be safely shared between threads, mutable
4033 portions may not be safely shared, but several mechanisms for effectively-safe
4034 sharing of mutable values, built on unsafe code but enforcing a safe locking
4035 discipline, exist in the standard library.
4037 Allocations in the stack consist of *variables*, and allocations in the heap
4040 ### Memory allocation and lifetime
4042 The _items_ of a program are those functions, modules and types that have their
4043 value calculated at compile-time and stored uniquely in the memory image of the
4044 rust process. Items are neither dynamically allocated nor freed.
4046 The _heap_ is a general term that describes boxes. The lifetime of an
4047 allocation in the heap depends on the lifetime of the box values pointing to
4048 it. Since box values may themselves be passed in and out of frames, or stored
4049 in the heap, heap allocations may outlive the frame they are allocated within.
4050 An allocation in the heap is guaranteed to reside at a single location in the
4051 heap for the whole lifetime of the allocation - it will never be relocated as
4052 a result of moving a box value.
4054 ### Memory ownership
4056 When a stack frame is exited, its local allocations are all released, and its
4057 references to boxes are dropped.
4061 A _variable_ is a component of a stack frame, either a named function parameter,
4062 an anonymous [temporary](#lvalues-rvalues-and-temporaries), or a named local
4065 A _local variable_ (or *stack-local* allocation) holds a value directly,
4066 allocated within the stack's memory. The value is a part of the stack frame.
4068 Local variables are immutable unless declared otherwise like: `let mut x = ...`.
4070 Function parameters are immutable unless declared with `mut`. The `mut` keyword
4071 applies only to the following parameter (so `|mut x, y|` and `fn f(mut x:
4072 Box<i32>, y: Box<i32>)` declare one mutable variable `x` and one immutable
4075 Methods that take either `self` or `Box<Self>` can optionally place them in a
4076 mutable variable by prefixing them with `mut` (similar to regular arguments):
4079 trait Changer: Sized {
4080 fn change(mut self) {}
4081 fn modify(mut self: Box<Self>) {}
4085 Local variables are not initialized when allocated; the entire frame worth of
4086 local variables are allocated at once, on frame-entry, in an uninitialized
4087 state. Subsequent statements within a function may or may not initialize the
4088 local variables. Local variables can be used only after they have been
4089 initialized; this is enforced by the compiler.
4093 The Rust compiler supports various methods to link crates together both
4094 statically and dynamically. This section will explore the various methods to
4095 link Rust crates together, and more information about native libraries can be
4096 found in the [FFI section of the book][ffi].
4098 In one session of compilation, the compiler can generate multiple artifacts
4099 through the usage of either command line flags or the `crate_type` attribute.
4100 If one or more command line flags are specified, all `crate_type` attributes will
4101 be ignored in favor of only building the artifacts specified by command line.
4103 * `--crate-type=bin`, `#[crate_type = "bin"]` - A runnable executable will be
4104 produced. This requires that there is a `main` function in the crate which
4105 will be run when the program begins executing. This will link in all Rust and
4106 native dependencies, producing a distributable binary.
4108 * `--crate-type=lib`, `#[crate_type = "lib"]` - A Rust library will be produced.
4109 This is an ambiguous concept as to what exactly is produced because a library
4110 can manifest itself in several forms. The purpose of this generic `lib` option
4111 is to generate the "compiler recommended" style of library. The output library
4112 will always be usable by rustc, but the actual type of library may change from
4113 time-to-time. The remaining output types are all different flavors of
4114 libraries, and the `lib` type can be seen as an alias for one of them (but the
4115 actual one is compiler-defined).
4117 * `--crate-type=dylib`, `#[crate_type = "dylib"]` - A dynamic Rust library will
4118 be produced. This is different from the `lib` output type in that this forces
4119 dynamic library generation. The resulting dynamic library can be used as a
4120 dependency for other libraries and/or executables. This output type will
4121 create `*.so` files on linux, `*.dylib` files on osx, and `*.dll` files on
4124 * `--crate-type=staticlib`, `#[crate_type = "staticlib"]` - A static system
4125 library will be produced. This is different from other library outputs in that
4126 the Rust compiler will never attempt to link to `staticlib` outputs. The
4127 purpose of this output type is to create a static library containing all of
4128 the local crate's code along with all upstream dependencies. The static
4129 library is actually a `*.a` archive on linux and osx and a `*.lib` file on
4130 windows. This format is recommended for use in situations such as linking
4131 Rust code into an existing non-Rust application because it will not have
4132 dynamic dependencies on other Rust code.
4134 * `--crate-type=cdylib`, `#[crate_type = "cdylib"]` - A dynamic system
4135 library will be produced. This is used when compiling Rust code as
4136 a dynamic library to be loaded from another language. This output type will
4137 create `*.so` files on Linux, `*.dylib` files on OSX, and `*.dll` files on
4140 * `--crate-type=rlib`, `#[crate_type = "rlib"]` - A "Rust library" file will be
4141 produced. This is used as an intermediate artifact and can be thought of as a
4142 "static Rust library". These `rlib` files, unlike `staticlib` files, are
4143 interpreted by the Rust compiler in future linkage. This essentially means
4144 that `rustc` will look for metadata in `rlib` files like it looks for metadata
4145 in dynamic libraries. This form of output is used to produce statically linked
4146 executables as well as `staticlib` outputs.
4148 * `--crate-type=proc-macro`, `#[crate_type = "proc-macro"]` - The output
4149 produced is not specified, but if a `-L` path is provided to it then the
4150 compiler will recognize the output artifacts as a macro and it can be loaded
4151 for a program. If a crate is compiled with the `proc-macro` crate type it
4152 will forbid exporting any items in the crate other than those functions
4153 tagged `#[proc_macro_derive]` and those functions must also be placed at the
4154 crate root. Finally, the compiler will automatically set the
4155 `cfg(proc_macro)` annotation whenever any crate type of a compilation is the
4156 `proc-macro` crate type.
4158 Note that these outputs are stackable in the sense that if multiple are
4159 specified, then the compiler will produce each form of output at once without
4160 having to recompile. However, this only applies for outputs specified by the
4161 same method. If only `crate_type` attributes are specified, then they will all
4162 be built, but if one or more `--crate-type` command line flags are specified,
4163 then only those outputs will be built.
4165 With all these different kinds of outputs, if crate A depends on crate B, then
4166 the compiler could find B in various different forms throughout the system. The
4167 only forms looked for by the compiler, however, are the `rlib` format and the
4168 dynamic library format. With these two options for a dependent library, the
4169 compiler must at some point make a choice between these two formats. With this
4170 in mind, the compiler follows these rules when determining what format of
4171 dependencies will be used:
4173 1. If a static library is being produced, all upstream dependencies are
4174 required to be available in `rlib` formats. This requirement stems from the
4175 reason that a dynamic library cannot be converted into a static format.
4177 Note that it is impossible to link in native dynamic dependencies to a static
4178 library, and in this case warnings will be printed about all unlinked native
4179 dynamic dependencies.
4181 2. If an `rlib` file is being produced, then there are no restrictions on what
4182 format the upstream dependencies are available in. It is simply required that
4183 all upstream dependencies be available for reading metadata from.
4185 The reason for this is that `rlib` files do not contain any of their upstream
4186 dependencies. It wouldn't be very efficient for all `rlib` files to contain a
4187 copy of `libstd.rlib`!
4189 3. If an executable is being produced and the `-C prefer-dynamic` flag is not
4190 specified, then dependencies are first attempted to be found in the `rlib`
4191 format. If some dependencies are not available in an rlib format, then
4192 dynamic linking is attempted (see below).
4194 4. If a dynamic library or an executable that is being dynamically linked is
4195 being produced, then the compiler will attempt to reconcile the available
4196 dependencies in either the rlib or dylib format to create a final product.
4198 A major goal of the compiler is to ensure that a library never appears more
4199 than once in any artifact. For example, if dynamic libraries B and C were
4200 each statically linked to library A, then a crate could not link to B and C
4201 together because there would be two copies of A. The compiler allows mixing
4202 the rlib and dylib formats, but this restriction must be satisfied.
4204 The compiler currently implements no method of hinting what format a library
4205 should be linked with. When dynamically linking, the compiler will attempt to
4206 maximize dynamic dependencies while still allowing some dependencies to be
4207 linked in via an rlib.
4209 For most situations, having all libraries available as a dylib is recommended
4210 if dynamically linking. For other situations, the compiler will emit a
4211 warning if it is unable to determine which formats to link each library with.
4213 In general, `--crate-type=bin` or `--crate-type=lib` should be sufficient for
4214 all compilation needs, and the other options are just available if more
4215 fine-grained control is desired over the output format of a Rust crate.
4219 Unsafe operations are those that potentially violate the memory-safety
4220 guarantees of Rust's static semantics.
4222 The following language level features cannot be used in the safe subset of
4225 - Dereferencing a [raw pointer](#pointer-types).
4226 - Reading or writing a [mutable static variable](#mutable-statics).
4227 - Calling an unsafe function (including an intrinsic or foreign function).
4231 Unsafe functions are functions that are not safe in all contexts and/or for all
4232 possible inputs. Such a function must be prefixed with the keyword `unsafe` and
4233 can only be called from an `unsafe` block or another `unsafe` function.
4237 A block of code can be prefixed with the `unsafe` keyword, to permit calling
4238 `unsafe` functions or dereferencing raw pointers within a safe function.
4240 When a programmer has sufficient conviction that a sequence of potentially
4241 unsafe operations is actually safe, they can encapsulate that sequence (taken
4242 as a whole) within an `unsafe` block. The compiler will consider uses of such
4243 code safe, in the surrounding context.
4245 Unsafe blocks are used to wrap foreign libraries, make direct use of hardware
4246 or implement features not directly present in the language. For example, Rust
4247 provides the language features necessary to implement memory-safe concurrency
4248 in the language but the implementation of threads and message passing is in the
4251 Rust's type system is a conservative approximation of the dynamic safety
4252 requirements, so in some cases there is a performance cost to using safe code.
4253 For example, a doubly-linked list is not a tree structure and can only be
4254 represented with reference-counted pointers in safe code. By using `unsafe`
4255 blocks to represent the reverse links as raw pointers, it can be implemented
4258 ## Behavior considered undefined
4260 The following is a list of behavior which is forbidden in all Rust code,
4261 including within `unsafe` blocks and `unsafe` functions. Type checking provides
4262 the guarantee that these issues are never caused by safe code.
4265 * Dereferencing a null/dangling raw pointer
4266 * Reads of [undef](http://llvm.org/docs/LangRef.html#undefined-values)
4267 (uninitialized) memory
4268 * Breaking the [pointer aliasing
4269 rules](http://llvm.org/docs/LangRef.html#pointer-aliasing-rules)
4270 with raw pointers (a subset of the rules used by C)
4271 * `&mut T` and `&T` follow LLVM’s scoped [noalias] model, except if the `&T`
4272 contains an `UnsafeCell<U>`. Unsafe code must not violate these aliasing
4274 * Mutating non-mutable data (that is, data reached through a shared reference or
4275 data owned by a `let` binding), unless that data is contained within an `UnsafeCell<U>`.
4276 * Invoking undefined behavior via compiler intrinsics:
4277 * Indexing outside of the bounds of an object with `std::ptr::offset`
4278 (`offset` intrinsic), with
4279 the exception of one byte past the end which is permitted.
4280 * Using `std::ptr::copy_nonoverlapping_memory` (`memcpy32`/`memcpy64`
4281 intrinsics) on overlapping buffers
4282 * Invalid values in primitive types, even in private fields/locals:
4283 * Dangling/null references or boxes
4284 * A value other than `false` (0) or `true` (1) in a `bool`
4285 * A discriminant in an `enum` not included in the type definition
4286 * A value in a `char` which is a surrogate or above `char::MAX`
4287 * Non-UTF-8 byte sequences in a `str`
4288 * Unwinding into Rust from foreign code or unwinding from Rust into foreign
4289 code. Rust's failure system is not compatible with exception handling in
4290 other languages. Unwinding must be caught and handled at FFI boundaries.
4292 [noalias]: http://llvm.org/docs/LangRef.html#noalias
4294 ## Behavior not considered unsafe
4296 This is a list of behavior not considered *unsafe* in Rust terms, but that may
4300 * Leaks of memory and other resources
4301 * Exiting without calling destructors
4303 - Overflow is considered "unexpected" behavior and is always user-error,
4304 unless the `wrapping` primitives are used. In non-optimized builds, the compiler
4305 will insert debug checks that panic on overflow, but in optimized builds overflow
4306 instead results in wrapped values. See [RFC 560] for the rationale and more details.
4308 [RFC 560]: https://github.com/rust-lang/rfcs/blob/master/text/0560-integer-overflow.md
4310 # Appendix: Influences
4312 Rust is not a particularly original language, with design elements coming from
4313 a wide range of sources. Some of these are listed below (including elements
4314 that have since been removed):
4316 * SML, OCaml: algebraic data types, pattern matching, type inference,
4317 semicolon statement separation
4318 * C++: references, RAII, smart pointers, move semantics, monomorphization,
4320 * ML Kit, Cyclone: region based memory management
4321 * Haskell (GHC): typeclasses, type families
4322 * Newsqueak, Alef, Limbo: channels, concurrency
4323 * Erlang: message passing, thread failure, ~~linked thread failure~~,
4324 ~~lightweight concurrency~~
4325 * Swift: optional bindings
4326 * Scheme: hygienic macros
4328 * Ruby: ~~block syntax~~
4329 * NIL, Hermes: ~~typestate~~
4330 * [Unicode Annex #31](http://www.unicode.org/reports/tr31/): identifier and
4333 [ffi]: book/ffi.html
4334 [plugin]: book/compiler-plugins.html
4335 [procedural macros]: book/procedural-macros.html