5 This document is the primary reference for the Rust programming language. It
6 provides three kinds of material:
8 - Chapters that formally define the language grammar and, for each
9 construct, informally describe its semantics and give examples of its
11 - Chapters that informally describe the memory model, concurrency model,
12 runtime services, linkage model and debugging facilities.
13 - Appendix chapters providing rationale and references to languages that
14 influenced the design.
16 This document does not serve as an introduction to the language. Background
17 familiarity with the language is assumed. A separate [book] is available to
18 help acquire such background familiarity.
20 This document also does not serve as a reference to the [standard] library
21 included in the language distribution. Those libraries are documented
22 separately by extracting documentation attributes from their source code. Many
23 of the features that one might expect to be language features are library
24 features in Rust, so what you're looking for may be there, not here.
26 [book]: book/index.html
27 [standard]: std/index.html
31 Rust's grammar is defined over Unicode codepoints, each conventionally denoted
32 `U+XXXX`, for 4 or more hexadecimal digits `X`. _Most_ of Rust's grammar is
33 confined to the ASCII range of Unicode, and is described in this document by a
34 dialect of Extended Backus-Naur Form (EBNF), specifically a dialect of EBNF
35 supported by common automated LL(k) parsing tools such as `llgen`, rather than
36 the dialect given in ISO 14977. The dialect can be defined self-referentially
41 rule : nonterminal ':' productionrule ';' ;
42 productionrule : production [ '|' production ] * ;
44 term : element repeats ;
45 element : LITERAL | IDENTIFIER | '[' productionrule ']' ;
46 repeats : [ '*' | '+' ] NUMBER ? | NUMBER ? | '?' ;
51 - Whitespace in the grammar is ignored.
52 - Square brackets are used to group rules.
53 - `LITERAL` is a single printable ASCII character, or an escaped hexadecimal
54 ASCII code of the form `\xQQ`, in single quotes, denoting the corresponding
55 Unicode codepoint `U+00QQ`.
56 - `IDENTIFIER` is a nonempty string of ASCII letters and underscores.
57 - The `repeat` forms apply to the adjacent `element`, and are as follows:
58 - `?` means zero or one repetition
59 - `*` means zero or more repetitions
60 - `+` means one or more repetitions
61 - NUMBER trailing a repeat symbol gives a maximum repetition count
62 - NUMBER on its own gives an exact repetition count
64 This EBNF dialect should hopefully be familiar to many readers.
66 ## Unicode productions
68 A few productions in Rust's grammar permit Unicode codepoints outside the ASCII
69 range. We define these productions in terms of character properties specified
70 in the Unicode standard, rather than in terms of ASCII-range codepoints. The
71 section [Special Unicode Productions](#special-unicode-productions) lists these
74 ## String table productions
76 Some rules in the grammar — notably [unary
77 operators](#unary-operator-expressions), [binary
78 operators](#binary-operator-expressions), and [keywords](#keywords) — are
79 given in a simplified form: as a listing of a table of unquoted, printable
80 whitespace-separated strings. These cases form a subset of the rules regarding
81 the [token](#tokens) rule, and are assumed to be the result of a
82 lexical-analysis phase feeding the parser, driven by a DFA, operating over the
83 disjunction of all such string table entries.
85 When such a string enclosed in double-quotes (`"`) occurs inside the grammar,
86 it is an implicit reference to a single member of such a string table
87 production. See [tokens](#tokens) for more information.
93 Rust input is interpreted as a sequence of Unicode codepoints encoded in UTF-8.
94 Most Rust grammar rules are defined in terms of printable ASCII-range
95 codepoints, but a small number are defined in terms of Unicode properties or
96 explicit codepoint lists. [^inputformat]
98 [^inputformat]: Substitute definitions for the special Unicode productions are
99 provided to the grammar verifier, restricted to ASCII range, when verifying the
100 grammar in this document.
102 ## Special Unicode Productions
104 The following productions in the Rust grammar are defined in terms of Unicode
105 properties: `ident`, `non_null`, `non_star`, `non_eol`, `non_slash_or_star`,
106 `non_single_quote` and `non_double_quote`.
110 The `ident` production is any nonempty Unicode string of the following form:
112 - The first character has property `XID_start`
113 - The remaining characters have property `XID_continue`
115 that does _not_ occur in the set of [keywords](#keywords).
117 > **Note**: `XID_start` and `XID_continue` as character properties cover the
118 > character ranges used to form the more familiar C and Java language-family
121 ### Delimiter-restricted productions
123 Some productions are defined by exclusion of particular Unicode characters:
125 - `non_null` is any single Unicode character aside from `U+0000` (null)
126 - `non_eol` is `non_null` restricted to exclude `U+000A` (`'\n'`)
127 - `non_star` is `non_null` restricted to exclude `U+002A` (`*`)
128 - `non_slash_or_star` is `non_null` restricted to exclude `U+002F` (`/`) and `U+002A` (`*`)
129 - `non_single_quote` is `non_null` restricted to exclude `U+0027` (`'`)
130 - `non_double_quote` is `non_null` restricted to exclude `U+0022` (`"`)
135 comment : block_comment | line_comment ;
136 block_comment : "/*" block_comment_body * "*/" ;
137 block_comment_body : [block_comment | character] * ;
138 line_comment : "//" non_eol * ;
141 Comments in Rust code follow the general C++ style of line and block-comment
142 forms. Nested block comments are supported.
144 Line comments beginning with exactly _three_ slashes (`///`), and block
145 comments beginning with exactly one repeated asterisk in the block-open
146 sequence (`/**`), are interpreted as a special syntax for `doc`
147 [attributes](#attributes). That is, they are equivalent to writing
148 `#[doc="..."]` around the body of the comment (this includes the comment
149 characters themselves, ie `/// Foo` turns into `#[doc="/// Foo"]`).
151 `//!` comments apply to the parent of the comment, rather than the item that
152 follows. `//!` comments are usually used to display information on the crate
155 Non-doc comments are interpreted as a form of whitespace.
160 whitespace_char : '\x20' | '\x09' | '\x0a' | '\x0d' ;
161 whitespace : [ whitespace_char | comment ] + ;
164 The `whitespace_char` production is any nonempty Unicode string consisting of
165 any of the following Unicode characters: `U+0020` (space, `' '`), `U+0009`
166 (tab, `'\t'`), `U+000A` (LF, `'\n'`), `U+000D` (CR, `'\r'`).
168 Rust is a "free-form" language, meaning that all forms of whitespace serve only
169 to separate _tokens_ in the grammar, and have no semantic significance.
171 A Rust program has identical meaning if each whitespace element is replaced
172 with any other legal whitespace element, such as a single space character.
177 simple_token : keyword | unop | binop ;
178 token : simple_token | ident | literal | symbol | whitespace token ;
181 Tokens are primitive productions in the grammar defined by regular
182 (non-recursive) languages. "Simple" tokens are given in [string table
183 production](#string-table-productions) form, and occur in the rest of the
184 grammar as double-quoted strings. Other tokens have exact rules given.
188 <p id="keyword-table-marker"></p>
191 |----------|----------|----------|----------|---------|
192 | abstract | alignof | as | be | box |
193 | break | const | continue | crate | do |
194 | else | enum | extern | false | final |
195 | fn | for | if | impl | in |
196 | let | loop | macro | match | mod |
197 | move | mut | offsetof | override | priv |
198 | pub | pure | ref | return | sizeof |
199 | static | self | struct | super | true |
200 | trait | type | typeof | unsafe | unsized |
201 | use | virtual | where | while | yield |
204 Each of these keywords has special meaning in its grammar, and all of them are
205 excluded from the `ident` rule.
207 Note that some of these keywords are reserved, and do not currently do
212 A literal is an expression consisting of a single token, rather than a sequence
213 of tokens, that immediately and directly denotes the value it evaluates to,
214 rather than referring to it by name or some other evaluation rule. A literal is
215 a form of constant expression, so is evaluated (primarily) at compile time.
219 literal : [ string_lit | char_lit | byte_string_lit | byte_lit | num_lit ] lit_suffix ?;
222 The optional suffix is only used for certain numeric literals, but is
223 reserved for future extension, that is, the above gives the lexical
224 grammar, but a Rust parser will reject everything but the 12 special
225 cases mentioned in [Number literals](#number-literals) below.
229 ##### Characters and strings
231 | | Example | Number of `#` pairs allowed | Available characters | Escapes | Equivalent to |
232 |---|---------|-----------------------------|----------------------|---------|---------------|
233 | [Character](#character-literals) | `'H'` | `N/A` | All unicode | `\'` & [Byte escapes](#byte-escapes) & [Unicode escapes](#unicode-escapes) | `N/A` |
234 | [String](#string-literals) | `"hello"` | `N/A` | All unicode | `\"` & [Byte escapes](#byte-escapes) & [Unicode escapes](#unicode-escapes) | `N/A` |
235 | [Raw](#raw-string-literals) | `r##"hello"##` | `0...` | All unicode | `N/A` | `N/A` |
236 | [Byte](#byte-literals) | `b'H'` | `N/A` | All ASCII | `\'` & [Byte escapes](#byte-escapes) | `u8` |
237 | [Byte string](#byte-string-literals) | `b"hello"` | `N/A` | All ASCII | `\"` & [Byte escapes](#byte-escapes) | `&'static [u8]` |
238 | [Raw byte string](#raw-byte-string-literals) | `br##"hello"##` | `0...` | All ASCII | `N/A` | `&'static [u8]` (unsure...not stated) |
244 | `\x7F` | 8-bit character code (exactly 2 digits) |
246 | `\r` | Carriage return |
250 ##### Unicode escapes
253 | `\u7FFF` | 16-bit character code (exactly 4 digits) |
254 | `\U7EEEFFFF` | 32-bit character code (exactly 8 digits) |
258 | [Number literals](#number-literals)`*` | Example | Exponentiation | Suffixes |
259 |----------------------------------------|---------|----------------|----------|
260 | Decimal integer | `98_222is` | `N/A` | Integer suffixes |
261 | Hex integer | `0xffis` | `N/A` | Integer suffixes |
262 | Octal integer | `0o77is` | `N/A` | Integer suffixes |
263 | Binary integer | `0b1111_0000is` | `N/A` | Integer suffixes |
264 | Floating-point | `123.0E+77f64` | `Optional` | Floating-point suffixes |
266 `*` All number literals allow `_` as a visual separator: `1_234.0E+18f64`
269 | Integer | Floating-point |
270 |---------|----------------|
271 | `is` (`isize`), `us` (`usize`), `u8`, `i8`, `u16`, `i16`, `u32`, `i32`, `u64`, `i64` | `f32`, `f64` |
273 #### Character and string literals
276 char_lit : '\x27' char_body '\x27' ;
277 string_lit : '"' string_body * '"' | 'r' raw_string ;
279 char_body : non_single_quote
280 | '\x5c' [ '\x27' | common_escape | unicode_escape ] ;
282 string_body : non_double_quote
283 | '\x5c' [ '\x22' | common_escape | unicode_escape ] ;
284 raw_string : '"' raw_string_body '"' | '#' raw_string '#' ;
286 common_escape : '\x5c'
287 | 'n' | 'r' | 't' | '0'
289 unicode_escape : 'u' hex_digit 4
292 hex_digit : 'a' | 'b' | 'c' | 'd' | 'e' | 'f'
293 | 'A' | 'B' | 'C' | 'D' | 'E' | 'F'
295 oct_digit : '0' | '1' | '2' | '3' | '4' | '5' | '6' | '7' ;
296 dec_digit : '0' | nonzero_dec ;
297 nonzero_dec: '1' | '2' | '3' | '4'
298 | '5' | '6' | '7' | '8' | '9' ;
301 ##### Character literals
303 A _character literal_ is a single Unicode character enclosed within two
304 `U+0027` (single-quote) characters, with the exception of `U+0027` itself,
305 which must be _escaped_ by a preceding U+005C character (`\`).
307 ##### String literals
309 A _string literal_ is a sequence of any Unicode characters enclosed within two
310 `U+0022` (double-quote) characters, with the exception of `U+0022` itself,
311 which must be _escaped_ by a preceding `U+005C` character (`\`), or a _raw
314 ##### Character escapes
316 Some additional _escapes_ are available in either character or non-raw string
317 literals. An escape starts with a `U+005C` (`\`) and continues with one of the
320 * An _8-bit codepoint escape_ escape starts with `U+0078` (`x`) and is
321 followed by exactly two _hex digits_. It denotes the Unicode codepoint
322 equal to the provided hex value.
323 * A _16-bit codepoint escape_ starts with `U+0075` (`u`) and is followed
324 by exactly four _hex digits_. It denotes the Unicode codepoint equal to
325 the provided hex value.
326 * A _32-bit codepoint escape_ starts with `U+0055` (`U`) and is followed
327 by exactly eight _hex digits_. It denotes the Unicode codepoint equal to
328 the provided hex value.
329 * A _whitespace escape_ is one of the characters `U+006E` (`n`), `U+0072`
330 (`r`), or `U+0074` (`t`), denoting the unicode values `U+000A` (LF),
331 `U+000D` (CR) or `U+0009` (HT) respectively.
332 * The _backslash escape_ is the character `U+005C` (`\`) which must be
333 escaped in order to denote *itself*.
335 ##### Raw string literals
337 Raw string literals do not process any escapes. They start with the character
338 `U+0072` (`r`), followed by zero or more of the character `U+0023` (`#`) and a
339 `U+0022` (double-quote) character. The _raw string body_ is not defined in the
340 EBNF grammar above: it can contain any sequence of Unicode characters and is
341 terminated only by another `U+0022` (double-quote) character, followed by the
342 same number of `U+0023` (`#`) characters that preceded the opening `U+0022`
343 (double-quote) character.
345 All Unicode characters contained in the raw string body represent themselves,
346 the characters `U+0022` (double-quote) (except when followed by at least as
347 many `U+0023` (`#`) characters as were used to start the raw string literal) or
348 `U+005C` (`\`) do not have any special meaning.
350 Examples for string literals:
353 "foo"; r"foo"; // foo
354 "\"foo\""; r#""foo""#; // "foo"
357 r##"foo #"# bar"##; // foo #"# bar
359 "\x52"; "R"; r"R"; // R
360 "\\x52"; r"\x52"; // \x52
363 #### Byte and byte string literals
366 byte_lit : "b\x27" byte_body '\x27' ;
367 byte_string_lit : "b\x22" string_body * '\x22' | "br" raw_byte_string ;
369 byte_body : ascii_non_single_quote
370 | '\x5c' [ '\x27' | common_escape ] ;
372 byte_string_body : ascii_non_double_quote
373 | '\x5c' [ '\x22' | common_escape ] ;
374 raw_byte_string : '"' raw_byte_string_body '"' | '#' raw_byte_string '#' ;
380 A _byte literal_ is a single ASCII character (in the `U+0000` to `U+007F`
381 range) enclosed within two `U+0027` (single-quote) characters, with the
382 exception of `U+0027` itself, which must be _escaped_ by a preceding U+005C
383 character (`\`), or a single _escape_. It is equivalent to a `u8` unsigned
384 8-bit integer _number literal_.
386 ##### Byte string literals
388 A _byte string literal_ is a sequence of ASCII characters and _escapes_
389 enclosed within two `U+0022` (double-quote) characters, with the exception of
390 `U+0022` itself, which must be _escaped_ by a preceding `U+005C` character
391 (`\`), or a _raw byte string literal_. It is equivalent to a `&'static [u8]`
392 borrowed array of unsigned 8-bit integers.
394 Some additional _escapes_ are available in either byte or non-raw byte string
395 literals. An escape starts with a `U+005C` (`\`) and continues with one of the
398 * An _byte escape_ escape starts with `U+0078` (`x`) and is
399 followed by exactly two _hex digits_. It denotes the byte
400 equal to the provided hex value.
401 * A _whitespace escape_ is one of the characters `U+006E` (`n`), `U+0072`
402 (`r`), or `U+0074` (`t`), denoting the bytes values `0x0A` (ASCII LF),
403 `0x0D` (ASCII CR) or `0x09` (ASCII HT) respectively.
404 * The _backslash escape_ is the character `U+005C` (`\`) which must be
405 escaped in order to denote its ASCII encoding `0x5C`.
407 ##### Raw byte string literals
409 Raw byte string literals do not process any escapes. They start with the
410 character `U+0062` (`b`), followed by `U+0072` (`r`), followed by zero or more
411 of the character `U+0023` (`#`), and a `U+0022` (double-quote) character. The
412 _raw string body_ is not defined in the EBNF grammar above: it can contain any
413 sequence of ASCII characters and is terminated only by another `U+0022`
414 (double-quote) character, followed by the same number of `U+0023` (`#`)
415 characters that preceded the opening `U+0022` (double-quote) character. A raw
416 byte string literal can not contain any non-ASCII byte.
418 All characters contained in the raw string body represent their ASCII encoding,
419 the characters `U+0022` (double-quote) (except when followed by at least as
420 many `U+0023` (`#`) characters as were used to start the raw string literal) or
421 `U+005C` (`\`) do not have any special meaning.
423 Examples for byte string literals:
426 b"foo"; br"foo"; // foo
427 b"\"foo\""; br#""foo""#; // "foo"
430 br##"foo #"# bar"##; // foo #"# bar
432 b"\x52"; b"R"; br"R"; // R
433 b"\\x52"; br"\x52"; // \x52
439 num_lit : nonzero_dec [ dec_digit | '_' ] * float_suffix ?
440 | '0' [ [ dec_digit | '_' ] * float_suffix ?
441 | 'b' [ '1' | '0' | '_' ] +
442 | 'o' [ oct_digit | '_' ] +
443 | 'x' [ hex_digit | '_' ] + ] ;
445 float_suffix : [ exponent | '.' dec_lit exponent ? ] ? ;
447 exponent : ['E' | 'e'] ['-' | '+' ] ? dec_lit ;
448 dec_lit : [ dec_digit | '_' ] + ;
451 A _number literal_ is either an _integer literal_ or a _floating-point
452 literal_. The grammar for recognizing the two kinds of literals is mixed.
454 ##### Integer literals
456 An _integer literal_ has one of four forms:
458 * A _decimal literal_ starts with a *decimal digit* and continues with any
459 mixture of *decimal digits* and _underscores_.
460 * A _hex literal_ starts with the character sequence `U+0030` `U+0078`
461 (`0x`) and continues as any mixture of hex digits and underscores.
462 * An _octal literal_ starts with the character sequence `U+0030` `U+006F`
463 (`0o`) and continues as any mixture of octal digits and underscores.
464 * A _binary literal_ starts with the character sequence `U+0030` `U+0062`
465 (`0b`) and continues as any mixture of binary digits and underscores.
467 Like any literal, an integer literal may be followed (immediately,
468 without any spaces) by an _integer suffix_, which forcibly sets the
469 type of the literal. There are 10 valid values for an integer suffix:
471 * The `is` and `us` suffixes give the literal type `isize` or `usize`,
473 * Each of the signed and unsigned machine types `u8`, `i8`,
474 `u16`, `i16`, `u32`, `i32`, `u64` and `i64`
475 give the literal the corresponding machine type.
477 The type of an _unsuffixed_ integer literal is determined by type inference.
478 If an integer type can be _uniquely_ determined from the surrounding program
479 context, the unsuffixed integer literal has that type. If the program context
480 underconstrains the type, it is considered a static type error; if the program
481 context overconstrains the type, it is also considered a static type error.
483 Examples of integer literals of various forms:
488 123_us; // type usize
490 0o70_i16; // type i16
491 0b1111_1111_1001_0000_i32; // type i32
494 ##### Floating-point literals
496 A _floating-point literal_ has one of two forms:
498 * A _decimal literal_ followed by a period character `U+002E` (`.`). This is
499 optionally followed by another decimal literal, with an optional _exponent_.
500 * A single _decimal literal_ followed by an _exponent_.
502 By default, a floating-point literal has a generic type, and, like integer
503 literals, the type must be uniquely determined from the context. There are two valid
504 _floating-point suffixes_, `f32` and `f64` (the 32-bit and 64-bit floating point
505 types), which explicitly determine the type of the literal.
507 Examples of floating-point literals of various forms:
510 123.0f64; // type f64
513 12E+99_f64; // type f64
514 let x: f64 = 2.; // type f64
517 This last example is different because it is not possible to use the suffix
518 syntax with a floating point literal ending in a period. `2.f64` would attempt
519 to call a method named `f64` on `2`.
521 #### Boolean literals
523 The two values of the boolean type are written `true` and `false`.
529 | '#' | '[' | ']' | '(' | ')' | '{' | '}'
533 Symbols are a general class of printable [token](#tokens) that play structural
534 roles in a variety of grammar productions. They are catalogued here for
535 completeness as the set of remaining miscellaneous printable tokens that do not
536 otherwise appear as [unary operators](#unary-operator-expressions), [binary
537 operators](#binary-operator-expressions), or [keywords](#keywords).
543 expr_path : [ "::" ] ident [ "::" expr_path_tail ] + ;
544 expr_path_tail : '<' type_expr [ ',' type_expr ] + '>'
547 type_path : ident [ type_path_tail ] + ;
548 type_path_tail : '<' type_expr [ ',' type_expr ] + '>'
552 A _path_ is a sequence of one or more path components _logically_ separated by
553 a namespace qualifier (`::`). If a path consists of only one component, it may
554 refer to either an [item](#items) or a [slot](#memory-slots) in a local control
555 scope. If a path has multiple components, it refers to an item.
557 Every item has a _canonical path_ within its crate, but the path naming an item
558 is only meaningful within a given crate. There is no global namespace across
559 crates; an item's canonical path merely identifies it within the crate.
561 Two examples of simple paths consisting of only identifier components:
568 Path components are usually [identifiers](#identifiers), but the trailing
569 component of a path may be an angle-bracket-enclosed list of type arguments. In
570 [expression](#expressions) context, the type argument list is given after a
571 final (`::`) namespace qualifier in order to disambiguate it from a relational
572 expression involving the less-than symbol (`<`). In type expression context,
573 the final namespace qualifier is omitted.
575 Two examples of paths with type arguments:
578 # struct HashMap<K, V>;
580 # fn id<T>(t: T) -> T { t }
581 type T = HashMap<i32,String>; // Type arguments used in a type expression
582 let x = id::<i32>(10); // Type arguments used in a call expression
586 Paths can be denoted with various leading qualifiers to change the meaning of
589 * Paths starting with `::` are considered to be global paths where the
590 components of the path start being resolved from the crate root. Each
591 identifier in the path must resolve to an item.
599 ::a::foo(); // call a's foo function
605 * Paths starting with the keyword `super` begin resolution relative to the
606 parent module. Each further identifier must resolve to an item.
614 super::a::foo(); // call a's foo function
620 * Paths starting with the keyword `self` begin resolution relative to the
621 current module. Each further identifier must resolve to an item.
633 A number of minor features of Rust are not central enough to have their own
634 syntax, and yet are not implementable as functions. Instead, they are given
635 names, and invoked through a consistent syntax: `name!(...)`. Examples include:
637 * `format!` : format data into a string
638 * `env!` : look up an environment variable's value at compile time
639 * `file!`: return the path to the file being compiled
640 * `stringify!` : pretty-print the Rust expression given as an argument
641 * `include!` : include the Rust expression in the given file
642 * `include_str!` : include the contents of the given file as a string
643 * `include_bytes!` : include the contents of the given file as a binary blob
644 * `error!`, `warn!`, `info!`, `debug!` : provide diagnostic information.
646 All of the above extensions are expressions with values.
648 Users of `rustc` can define new syntax extensions in two ways:
650 * [Compiler plugins](book/syntax-extensions.html) can include arbitrary
651 Rust code that manipulates syntax trees at compile time.
653 * [Macros](book/macros.html) define new syntax in a higher-level,
659 expr_macro_rules : "macro_rules" '!' ident '(' macro_rule * ')' ;
660 macro_rule : '(' matcher * ')' "=>" '(' transcriber * ')' ';' ;
661 matcher : '(' matcher * ')' | '[' matcher * ']'
662 | '{' matcher * '}' | '$' ident ':' ident
663 | '$' '(' matcher * ')' sep_token? [ '*' | '+' ]
664 | non_special_token ;
665 transcriber : '(' transcriber * ')' | '[' transcriber * ']'
666 | '{' transcriber * '}' | '$' ident
667 | '$' '(' transcriber * ')' sep_token? [ '*' | '+' ]
668 | non_special_token ;
671 `macro_rules` allows users to define syntax extension in a declarative way. We
672 call such extensions "macros by example" or simply "macros" — to be distinguished
673 from the "procedural macros" defined in [compiler plugins][plugin].
675 Currently, macros can expand to expressions, statements, items, or patterns.
677 (A `sep_token` is any token other than `*` and `+`. A `non_special_token` is
678 any token other than a delimiter or `$`.)
680 The macro expander looks up macro invocations by name, and tries each macro
681 rule in turn. It transcribes the first successful match. Matching and
682 transcription are closely related to each other, and we will describe them
687 The macro expander matches and transcribes every token that does not begin with
688 a `$` literally, including delimiters. For parsing reasons, delimiters must be
689 balanced, but they are otherwise not special.
691 In the matcher, `$` _name_ `:` _designator_ matches the nonterminal in the Rust
692 syntax named by _designator_. Valid designators are `item`, `block`, `stmt`,
693 `pat`, `expr`, `ty` (type), `ident`, `path`, `tt` (either side of the `=>`
694 in macro rules). In the transcriber, the designator is already known, and so
695 only the name of a matched nonterminal comes after the dollar sign.
697 In both the matcher and transcriber, the Kleene star-like operator indicates
698 repetition. The Kleene star operator consists of `$` and parens, optionally
699 followed by a separator token, followed by `*` or `+`. `*` means zero or more
700 repetitions, `+` means at least one repetition. The parens are not matched or
701 transcribed. On the matcher side, a name is bound to _all_ of the names it
702 matches, in a structure that mimics the structure of the repetition encountered
703 on a successful match. The job of the transcriber is to sort that structure
706 The rules for transcription of these repetitions are called "Macro By Example".
707 Essentially, one "layer" of repetition is discharged at a time, and all of them
708 must be discharged by the time a name is transcribed. Therefore, `( $( $i:ident
709 ),* ) => ( $i )` is an invalid macro, but `( $( $i:ident ),* ) => ( $( $i:ident
710 ),* )` is acceptable (if trivial).
712 When Macro By Example encounters a repetition, it examines all of the `$`
713 _name_ s that occur in its body. At the "current layer", they all must repeat
714 the same number of times, so ` ( $( $i:ident ),* ; $( $j:ident ),* ) => ( $(
715 ($i,$j) ),* )` is valid if given the argument `(a,b,c ; d,e,f)`, but not
716 `(a,b,c ; d,e)`. The repetition walks through the choices at that layer in
717 lockstep, so the former input transcribes to `( (a,d), (b,e), (c,f) )`.
719 Nested repetitions are allowed.
721 ### Parsing limitations
723 The parser used by the macro system is reasonably powerful, but the parsing of
724 Rust syntax is restricted in two ways:
726 1. The parser will always parse as much as possible. If it attempts to match
727 `$i:expr [ , ]` against `8 [ , ]`, it will attempt to parse `i` as an array
728 index operation and fail. Adding a separator can solve this problem.
729 2. The parser must have eliminated all ambiguity by the time it reaches a `$`
730 _name_ `:` _designator_. This requirement most often affects name-designator
731 pairs when they occur at the beginning of, or immediately after, a `$(...)*`;
732 requiring a distinctive token in front can solve the problem.
734 ## Syntax extensions useful for the macro author
736 * `log_syntax!` : print out the arguments at compile time
737 * `trace_macros!` : supply `true` or `false` to enable or disable macro expansion logging
738 * `stringify!` : turn the identifier argument into a string literal
739 * `concat!` : concatenates a comma-separated list of literals
740 * `concat_idents!` : create a new identifier by concatenating the arguments
742 # Crates and source files
744 Rust is a *compiled* language. Its semantics obey a *phase distinction*
745 between compile-time and run-time. Those semantic rules that have a *static
746 interpretation* govern the success or failure of compilation. We refer to
747 these rules as "static semantics". Semantic rules called "dynamic semantics"
748 govern the behavior of programs at run-time. A program that fails to compile
749 due to violation of a compile-time rule has no defined dynamic semantics; the
750 compiler should halt with an error report, and produce no executable artifact.
752 The compilation model centers on artifacts called _crates_. Each compilation
753 processes a single crate in source form, and if successful, produces a single
754 crate in binary form: either an executable or a library.[^cratesourcefile]
756 [^cratesourcefile]: A crate is somewhat analogous to an *assembly* in the
757 ECMA-335 CLI model, a *library* in the SML/NJ Compilation Manager, a *unit*
758 in the Owens and Flatt module system, or a *configuration* in Mesa.
760 A _crate_ is a unit of compilation and linking, as well as versioning,
761 distribution and runtime loading. A crate contains a _tree_ of nested
762 [module](#modules) scopes. The top level of this tree is a module that is
763 anonymous (from the point of view of paths within the module) and any item
764 within a crate has a canonical [module path](#paths) denoting its location
765 within the crate's module tree.
767 The Rust compiler is always invoked with a single source file as input, and
768 always produces a single output crate. The processing of that source file may
769 result in other source files being loaded as modules. Source files have the
772 A Rust source file describes a module, the name and location of which —
773 in the module tree of the current crate — are defined from outside the
774 source file: either by an explicit `mod_item` in a referencing source file, or
775 by the name of the crate itself.
777 Each source file contains a sequence of zero or more `item` definitions, and
778 may optionally begin with any number of `attributes` that apply to the
779 containing module. Attributes on the anonymous crate module define important
780 metadata that influences the behavior of the compiler.
783 # #![allow(unused_attribute)]
785 #![crate_name = "projx"]
787 // Specify the output type
788 #![crate_type = "lib"]
791 #![warn(non_camel_case_types)]
794 A crate that contains a `main` function can be compiled to an executable. If a
795 `main` function is present, its return type must be [`unit`](#primitive-types)
796 and it must take no arguments.
798 # Items and attributes
800 Crates contain [items](#items), each of which may have some number of
801 [attributes](#attributes) attached to it.
806 item : mod_item | fn_item | type_item | struct_item | enum_item
807 | static_item | trait_item | impl_item | extern_block ;
810 An _item_ is a component of a crate; some module items can be defined in crate
811 files, but most are defined in source files. Items are organized within a crate
812 by a nested set of [modules](#modules). Every crate has a single "outermost"
813 anonymous module; all further items within the crate have [paths](#paths)
814 within the module tree of the crate.
816 Items are entirely determined at compile-time, generally remain fixed during
817 execution, and may reside in read-only memory.
819 There are several kinds of item:
821 * [modules](#modules)
822 * [functions](#functions)
823 * [type definitions](#type-definitions)
824 * [structures](#structures)
825 * [enumerations](#enumerations)
826 * [static items](#static-items)
828 * [implementations](#implementations)
830 Some items form an implicit scope for the declaration of sub-items. In other
831 words, within a function or module, declarations of items can (in many cases)
832 be mixed with the statements, control blocks, and similar artifacts that
833 otherwise compose the item body. The meaning of these scoped items is the same
834 as if the item was declared outside the scope — it is still a static item
835 — except that the item's *path name* within the module namespace is
836 qualified by the name of the enclosing item, or is private to the enclosing
837 item (in the case of functions). The grammar specifies the exact locations in
838 which sub-item declarations may appear.
842 All items except modules may be *parameterized* by type. Type parameters are
843 given as a comma-separated list of identifiers enclosed in angle brackets
844 (`<...>`), after the name of the item and before its definition. The type
845 parameters of an item are considered "part of the name", not part of the type
846 of the item. A referencing [path](#paths) must (in principle) provide type
847 arguments as a list of comma-separated types enclosed within angle brackets, in
848 order to refer to the type-parameterized item. In practice, the type-inference
849 system can usually infer such argument types from context. There are no
850 general type-parametric types, only type-parametric items. That is, Rust has
851 no notion of type abstraction: there are no first-class "forall" types.
856 mod_item : "mod" ident ( ';' | '{' mod '}' );
857 mod : [ view_item | item ] * ;
860 A module is a container for zero or more [view items](#view-items) and zero or
861 more [items](#items). The view items manage the visibility of the items defined
862 within the module, as well as the visibility of names from outside the module
863 when referenced from inside the module.
865 A _module item_ is a module, surrounded in braces, named, and prefixed with the
866 keyword `mod`. A module item introduces a new, named module into the tree of
867 modules making up a crate. Modules can nest arbitrarily.
869 An example of a module:
873 type Complex = (f64, f64);
874 fn sin(f: f64) -> f64 {
878 fn cos(f: f64) -> f64 {
882 fn tan(f: f64) -> f64 {
889 Modules and types share the same namespace. Declaring a named type with
890 the same name as a module in scope is forbidden: that is, a type definition,
891 trait, struct, enumeration, or type parameter can't shadow the name of a module
892 in scope, or vice versa.
894 A module without a body is loaded from an external file, by default with the
895 same name as the module, plus the `.rs` extension. When a nested submodule is
896 loaded from an external file, it is loaded from a subdirectory path that
897 mirrors the module hierarchy.
900 // Load the `vec` module from `vec.rs`
904 // Load the `local_data` module from `thread/local_data.rs`
909 The directories and files used for loading external file modules can be
910 influenced with the `path` attribute.
913 #[path = "thread_files"]
915 // Load the `local_data` module from `thread_files/tls.rs`
924 view_item : extern_crate_decl | use_decl ;
927 A view item manages the namespace of a module. View items do not define new
928 items, but rather, simply change other items' visibility. There are two
931 * [`extern crate` declarations](#extern-crate-declarations)
932 * [`use` declarations](#use-declarations)
934 ##### Extern crate declarations
937 extern_crate_decl : "extern" "crate" crate_name
938 crate_name: ident | ( string_lit "as" ident )
941 An _`extern crate` declaration_ specifies a dependency on an external crate.
942 The external crate is then bound into the declaring scope as the `ident`
943 provided in the `extern_crate_decl`.
945 The external crate is resolved to a specific `soname` at compile time, and a
946 runtime linkage requirement to that `soname` is passed to the linker for
947 loading at runtime. The `soname` is resolved at compile time by scanning the
948 compiler's library path and matching the optional `crateid` provided as a
949 string literal against the `crateid` attributes that were declared on the
950 external crate when it was compiled. If no `crateid` is provided, a default
951 `name` attribute is assumed, equal to the `ident` given in the
954 Three examples of `extern crate` declarations:
959 extern crate std; // equivalent to: extern crate std as std;
961 extern crate "std" as ruststd; // linking to 'std' under another name
964 ##### Use declarations
967 use_decl : "pub" ? "use" [ path "as" ident
970 path_glob : ident [ "::" [ path_glob
972 | '{' path_item [ ',' path_item ] * '}' ;
974 path_item : ident | "self" ;
977 A _use declaration_ creates one or more local name bindings synonymous with
978 some other [path](#paths). Usually a `use` declaration is used to shorten the
979 path required to refer to a module item. These declarations may appear at the
980 top of [modules](#modules) and [blocks](#blocks).
982 > **Note**: Unlike in many languages,
983 > `use` declarations in Rust do *not* declare linkage dependency with external crates.
984 > Rather, [`extern crate` declarations](#extern-crate-declarations) declare linkage dependencies.
986 Use declarations support a number of convenient shortcuts:
988 * Rebinding the target name as a new local name, using the syntax `use p::q::r as x;`
989 * Simultaneously binding a list of paths differing only in their final element,
990 using the glob-like brace syntax `use a::b::{c,d,e,f};`
991 * Binding all paths matching a given prefix, using the asterisk wildcard syntax
993 * Simultaneously binding a list of paths differing only in their final element
994 and their immediate parent module, using the `self` keyword, such as
995 `use a::b::{self, c, d};`
997 An example of `use` declarations:
1000 use std::iter::range_step;
1001 use std::option::Option::{Some, None};
1002 use std::collections::hash_map::{self, HashMap};
1005 fn bar(map1: HashMap<String, usize>, map2: hash_map::HashMap<String, usize>){}
1008 // Equivalent to 'std::iter::range_step(0us, 10, 2);'
1009 range_step(0us, 10, 2);
1011 // Equivalent to 'foo(vec![std::option::Option::Some(1.0f64),
1012 // std::option::Option::None]);'
1013 foo(vec![Some(1.0f64), None]);
1015 // Both `hash_map` and `HashMap` are in scope.
1016 let map1 = HashMap::new();
1017 let map2 = hash_map::HashMap::new();
1022 Like items, `use` declarations are private to the containing module, by
1023 default. Also like items, a `use` declaration can be public, if qualified by
1024 the `pub` keyword. Such a `use` declaration serves to _re-export_ a name. A
1025 public `use` declaration can therefore _redirect_ some public name to a
1026 different target definition: even a definition with a private canonical path,
1027 inside a different module. If a sequence of such redirections form a cycle or
1028 cannot be resolved unambiguously, they represent a compile-time error.
1030 An example of re-exporting:
1035 pub use quux::foo::{bar, baz};
1044 In this example, the module `quux` re-exports two public names defined in
1047 Also note that the paths contained in `use` items are relative to the crate
1048 root. So, in the previous example, the `use` refers to `quux::foo::{bar,
1049 baz}`, and not simply to `foo::{bar, baz}`. This also means that top-level
1050 module declarations should be at the crate root if direct usage of the declared
1051 modules within `use` items is desired. It is also possible to use `self` and
1052 `super` at the beginning of a `use` item to refer to the current and direct
1053 parent modules respectively. All rules regarding accessing declared modules in
1054 `use` declarations applies to both module declarations and `extern crate`
1057 An example of what will and will not work for `use` items:
1060 # #![allow(unused_imports)]
1061 use foo::core::iter; // good: foo is at the root of the crate
1062 use foo::baz::foobaz; // good: foo is at the root of the crate
1067 use foo::core::iter; // good: foo is at crate root
1068 // use core::iter; // bad: native is not at the crate root
1069 use self::baz::foobaz; // good: self refers to module 'foo'
1070 use foo::bar::foobar; // good: foo is at crate root
1077 use super::bar::foobar; // good: super refers to module 'foo'
1087 A _function item_ defines a sequence of [statements](#statements) and an
1088 optional final [expression](#expressions), along with a name and a set of
1089 parameters. Functions are declared with the keyword `fn`. Functions declare a
1090 set of *input* [*slots*](#memory-slots) as parameters, through which the caller
1091 passes arguments into the function, and an *output* [*slot*](#memory-slots)
1092 through which the function passes results back to the caller.
1094 A function may also be copied into a first-class *value*, in which case the
1095 value has the corresponding [*function type*](#function-types), and can be used
1096 otherwise exactly as a function item (with a minor additional cost of calling
1097 the function indirectly).
1099 Every control path in a function logically ends with a `return` expression or a
1100 diverging expression. If the outermost block of a function has a
1101 value-producing expression in its final-expression position, that expression is
1102 interpreted as an implicit `return` expression applied to the final-expression.
1104 An example of a function:
1107 fn add(x: i32, y: i32) -> i32 {
1112 As with `let` bindings, function arguments are irrefutable patterns, so any
1113 pattern that is valid in a let binding is also valid as an argument.
1116 fn first((value, _): (i32, i32)) -> i32 { value }
1120 #### Generic functions
1122 A _generic function_ allows one or more _parameterized types_ to appear in its
1123 signature. Each type parameter must be explicitly declared, in an
1124 angle-bracket-enclosed, comma-separated list following the function name.
1127 fn iter<T>(seq: &[T], f: |T|) {
1128 for elt in seq.iter() { f(elt); }
1130 fn map<T, U>(seq: &[T], f: |T| -> U) -> Vec<U> {
1131 let mut acc = vec![];
1132 for elt in seq.iter() { acc.push(f(elt)); }
1137 Inside the function signature and body, the name of the type parameter can be
1138 used as a type name.
1140 When a generic function is referenced, its type is instantiated based on the
1141 context of the reference. For example, calling the `iter` function defined
1142 above on `[1, 2]` will instantiate type parameter `T` with `isize`, and require
1143 the closure parameter to have type `fn(isize)`.
1145 The type parameters can also be explicitly supplied in a trailing
1146 [path](#paths) component after the function name. This might be necessary if
1147 there is not sufficient context to determine the type parameters. For example,
1148 `mem::size_of::<u32>() == 4`.
1150 Since a parameter type is opaque to the generic function, the set of operations
1151 that can be performed on it is limited. Values of parameter type can only be
1155 fn id<T>(x: T) -> T { x }
1158 Similarly, [trait](#traits) bounds can be specified for type parameters to
1159 allow methods with that trait to be called on values of that type.
1163 Unsafe operations are those that potentially violate the memory-safety
1164 guarantees of Rust's static semantics.
1166 The following language level features cannot be used in the safe subset of
1169 - Dereferencing a [raw pointer](#pointer-types).
1170 - Reading or writing a [mutable static variable](#mutable-statics).
1171 - Calling an unsafe function (including an intrinsic or foreign function).
1173 ##### Unsafe functions
1175 Unsafe functions are functions that are not safe in all contexts and/or for all
1176 possible inputs. Such a function must be prefixed with the keyword `unsafe` and
1177 can only be called from an `unsafe` block or another `unsafe` function.
1181 A block of code can be prefixed with the `unsafe` keyword, to permit calling
1182 `unsafe` functions or dereferencing raw pointers within a safe function.
1184 When a programmer has sufficient conviction that a sequence of potentially
1185 unsafe operations is actually safe, they can encapsulate that sequence (taken
1186 as a whole) within an `unsafe` block. The compiler will consider uses of such
1187 code safe, in the surrounding context.
1189 Unsafe blocks are used to wrap foreign libraries, make direct use of hardware
1190 or implement features not directly present in the language. For example, Rust
1191 provides the language features necessary to implement memory-safe concurrency
1192 in the language but the implementation of threads and message passing is in the
1195 Rust's type system is a conservative approximation of the dynamic safety
1196 requirements, so in some cases there is a performance cost to using safe code.
1197 For example, a doubly-linked list is not a tree structure and can only be
1198 represented with reference-counted pointers in safe code. By using `unsafe`
1199 blocks to represent the reverse links as raw pointers, it can be implemented
1202 ##### Behavior considered undefined
1204 The following is a list of behavior which is forbidden in all Rust code,
1205 including within `unsafe` blocks and `unsafe` functions. Type checking provides
1206 the guarantee that these issues are never caused by safe code.
1209 * Dereferencing a null/dangling raw pointer
1210 * Mutating an immutable value/reference without `UnsafeCell`
1211 * Reads of [undef](http://llvm.org/docs/LangRef.html#undefined-values)
1212 (uninitialized) memory
1213 * Breaking the [pointer aliasing
1214 rules](http://llvm.org/docs/LangRef.html#pointer-aliasing-rules)
1215 with raw pointers (a subset of the rules used by C)
1216 * Invoking undefined behavior via compiler intrinsics:
1217 * Indexing outside of the bounds of an object with `std::ptr::offset`
1218 (`offset` intrinsic), with
1219 the exception of one byte past the end which is permitted.
1220 * Using `std::ptr::copy_nonoverlapping_memory` (`memcpy32`/`memcpy64`
1221 intrinsics) on overlapping buffers
1222 * Invalid values in primitive types, even in private fields/locals:
1223 * Dangling/null references or boxes
1224 * A value other than `false` (0) or `true` (1) in a `bool`
1225 * A discriminant in an `enum` not included in the type definition
1226 * A value in a `char` which is a surrogate or above `char::MAX`
1227 * Non-UTF-8 byte sequences in a `str`
1228 * Unwinding into Rust from foreign code or unwinding from Rust into foreign
1229 code. Rust's failure system is not compatible with exception handling in
1230 other languages. Unwinding must be caught and handled at FFI boundaries.
1232 ##### Behaviour not considered unsafe
1234 This is a list of behaviour not considered *unsafe* in Rust terms, but that may
1238 * Reading data from private fields (`std::repr`)
1239 * Leaks due to reference count cycles, even in the global heap
1240 * Exiting without calling destructors
1242 * Accessing/modifying the file system
1243 * Unsigned integer overflow (well-defined as wrapping)
1244 * Signed integer overflow (well-defined as two's complement representation
1247 #### Diverging functions
1249 A special kind of function can be declared with a `!` character where the
1250 output slot type would normally be. For example:
1253 fn my_err(s: &str) -> ! {
1259 We call such functions "diverging" because they never return a value to the
1260 caller. Every control path in a diverging function must end with a `panic!()` or
1261 a call to another diverging function on every control path. The `!` annotation
1262 does *not* denote a type. Rather, the result type of a diverging function is a
1263 special type called ⊥ ("bottom") that unifies with any type. Rust has no
1266 It might be necessary to declare a diverging function because as mentioned
1267 previously, the typechecker checks that every control path in a function ends
1268 with a [`return`](#return-expressions) or diverging expression. So, if `my_err`
1269 were declared without the `!` annotation, the following code would not
1273 # fn my_err(s: &str) -> ! { panic!() }
1275 fn f(i: i32) -> i32 {
1280 my_err("Bad number!");
1285 This will not compile without the `!` annotation on `my_err`, since the `else`
1286 branch of the conditional in `f` does not return an `i32`, as required by the
1287 signature of `f`. Adding the `!` annotation to `my_err` informs the
1288 typechecker that, should control ever enter `my_err`, no further type judgments
1289 about `f` need to hold, since control will never resume in any context that
1290 relies on those judgments. Thus the return type on `f` only needs to reflect
1291 the `if` branch of the conditional.
1293 #### Extern functions
1295 Extern functions are part of Rust's foreign function interface, providing the
1296 opposite functionality to [external blocks](#external-blocks). Whereas
1297 external blocks allow Rust code to call foreign code, extern functions with
1298 bodies defined in Rust code _can be called by foreign code_. They are defined
1299 in the same way as any other Rust function, except that they have the `extern`
1303 // Declares an extern fn, the ABI defaults to "C"
1304 extern fn new_i32() -> i32 { 0 }
1306 // Declares an extern fn with "stdcall" ABI
1307 extern "stdcall" fn new_i32_stdcall() -> i32 { 0 }
1310 Unlike normal functions, extern fns have an `extern "ABI" fn()`. This is the
1311 same type as the functions declared in an extern block.
1314 # extern fn new_i32() -> i32 { 0 }
1315 let fptr: extern "C" fn() -> i32 = new_i32;
1318 Extern functions may be called directly from Rust code as Rust uses large,
1319 contiguous stack segments like C.
1323 A _type alias_ defines a new name for an existing [type](#types). Type
1324 aliases are declared with the keyword `type`. Every value has a single,
1325 specific type; the type-specified aspects of a value include:
1327 * Whether the value is composed of sub-values or is indivisible.
1328 * Whether the value represents textual or numerical information.
1329 * Whether the value represents integral or floating-point information.
1330 * The sequence of memory operations required to access the value.
1331 * The [kind](#type-kinds) of the type.
1333 For example, the type `(u8, u8)` defines the set of immutable values that are
1334 composite pairs, each containing two unsigned 8-bit integers accessed by
1335 pattern-matching and laid out in memory with the `x` component preceding the
1339 type Point = (u8, u8);
1340 let p: Point = (41, 68);
1345 A _structure_ is a nominal [structure type](#structure-types) defined with the
1348 An example of a `struct` item and its use:
1351 struct Point {x: i32, y: i32}
1352 let p = Point {x: 10, y: 11};
1356 A _tuple structure_ is a nominal [tuple type](#tuple-types), also defined with
1357 the keyword `struct`. For example:
1360 struct Point(i32, i32);
1361 let p = Point(10, 11);
1362 let px: i32 = match p { Point(x, _) => x };
1365 A _unit-like struct_ is a structure without any fields, defined by leaving off
1366 the list of fields entirely. Such types will have a single value, just like
1367 the [unit value `()`](#unit-and-boolean-literals) of the unit type. For
1372 let c = [Cookie, Cookie, Cookie, Cookie];
1375 The precise memory layout of a structure is not specified. One can specify a
1376 particular layout using the [`repr` attribute](#ffi-attributes).
1380 An _enumeration_ is a simultaneous definition of a nominal [enumerated
1381 type](#enumerated-types) as well as a set of *constructors*, that can be used
1382 to create or pattern-match values of the corresponding enumerated type.
1384 Enumerations are declared with the keyword `enum`.
1386 An example of an `enum` item and its use:
1394 let mut a: Animal = Animal::Dog;
1398 Enumeration constructors can have either named or unnamed fields:
1401 # #![feature(struct_variant)]
1405 Cat { name: String, weight: f64 }
1408 let mut a: Animal = Animal::Dog("Cocoa".to_string(), 37.2);
1409 a = Animal::Cat { name: "Spotty".to_string(), weight: 2.7 };
1413 In this example, `Cat` is a _struct-like enum variant_,
1414 whereas `Dog` is simply called an enum variant.
1416 Enums have a discriminant. You can assign them explicitly:
1424 If a discriminant isn't assigned, they start at zero, and add one for each
1427 You can cast an enum to get this value:
1430 # enum Foo { Bar = 123 }
1431 let x = Foo::Bar as u32; // x is now 123u32
1434 This only works as long as none of the variants have data attached. If
1435 it were `Bar(i32)`, this is disallowed.
1440 const_item : "const" ident ':' type '=' expr ';' ;
1443 A *constant item* is a named _constant value_ which is not associated with a
1444 specific memory location in the program. Constants are essentially inlined
1445 wherever they are used, meaning that they are copied directly into the relevant
1446 context when used. References to the same constant are not necessarily
1447 guaranteed to refer to the same memory address.
1449 Constant values must not have destructors, and otherwise permit most forms of
1450 data. Constants may refer to the address of other constants, in which case the
1451 address will have the `static` lifetime. The compiler is, however, still at
1452 liberty to translate the constant many times, so the address referred to may not
1455 Constants must be explicitly typed. The type may be `bool`, `char`, a number, or
1456 a type derived from those primitive types. The derived types are references with
1457 the `static` lifetime, fixed-size arrays, tuples, enum variants, and structs.
1460 const BIT1: u32 = 1 << 0;
1461 const BIT2: u32 = 1 << 1;
1463 const BITS: [u32; 2] = [BIT1, BIT2];
1464 const STRING: &'static str = "bitstring";
1466 struct BitsNStrings<'a> {
1471 const BITS_N_STRINGS: BitsNStrings<'static> = BitsNStrings {
1480 static_item : "static" ident ':' type '=' expr ';' ;
1483 A *static item* is similar to a *constant*, except that it represents a precise
1484 memory location in the program. A static is never "inlined" at the usage site,
1485 and all references to it refer to the same memory location. Static items have
1486 the `static` lifetime, which outlives all other lifetimes in a Rust program.
1487 Static items may be placed in read-only memory if they do not contain any
1488 interior mutability.
1490 Statics may contain interior mutability through the `UnsafeCell` language item.
1491 All access to a static is safe, but there are a number of restrictions on
1494 * Statics may not contain any destructors.
1495 * The types of static values must ascribe to `Sync` to allow threadsafe access.
1496 * Statics may not refer to other statics by value, only by reference.
1497 * Constants cannot refer to statics.
1499 Constants should in general be preferred over statics, unless large amounts of
1500 data are being stored, or single-address and mutability properties are required.
1503 use std::sync::atomic::{AtomicUsize, Ordering, ATOMIC_USIZE_INIT};
1505 // Note that ATOMIC_USIZE_INIT is a *const*, but it may be used to initialize a
1506 // static. This static can be modified, so it is not placed in read-only memory.
1507 static COUNTER: AtomicUsize = ATOMIC_USIZE_INIT;
1509 // This table is a candidate to be placed in read-only memory.
1510 static TABLE: &'static [usize] = &[1, 2, 3, /* ... */];
1512 for slot in TABLE.iter() {
1513 println!("{}", slot);
1515 COUNTER.fetch_add(1, Ordering::SeqCst);
1518 #### Mutable statics
1520 If a static item is declared with the `mut` keyword, then it is allowed to
1521 be modified by the program. One of Rust's goals is to make concurrency bugs
1522 hard to run into, and this is obviously a very large source of race conditions
1523 or other bugs. For this reason, an `unsafe` block is required when either
1524 reading or writing a mutable static variable. Care should be taken to ensure
1525 that modifications to a mutable static are safe with respect to other threads
1526 running in the same process.
1528 Mutable statics are still very useful, however. They can be used with C
1529 libraries and can also be bound from C libraries (in an `extern` block).
1532 # fn atomic_add(_: &mut u32, _: u32) -> u32 { 2 }
1534 static mut LEVELS: u32 = 0;
1536 // This violates the idea of no shared state, and this doesn't internally
1537 // protect against races, so this function is `unsafe`
1538 unsafe fn bump_levels_unsafe1() -> u32 {
1544 // Assuming that we have an atomic_add function which returns the old value,
1545 // this function is "safe" but the meaning of the return value may not be what
1546 // callers expect, so it's still marked as `unsafe`
1547 unsafe fn bump_levels_unsafe2() -> u32 {
1548 return atomic_add(&mut LEVELS, 1);
1552 Mutable statics have the same restrictions as normal statics, except that the
1553 type of the value is not required to ascribe to `Sync`.
1557 A _trait_ describes a set of method types.
1559 Traits can include default implementations of methods, written in terms of some
1560 unknown [`self` type](#self-types); the `self` type may either be completely
1561 unspecified, or constrained by some other trait.
1563 Traits are implemented for specific types through separate
1564 [implementations](#implementations).
1567 # type Surface = i32;
1568 # type BoundingBox = i32;
1570 fn draw(&self, Surface);
1571 fn bounding_box(&self) -> BoundingBox;
1575 This defines a trait with two methods. All values that have
1576 [implementations](#implementations) of this trait in scope can have their
1577 `draw` and `bounding_box` methods called, using `value.bounding_box()`
1578 [syntax](#method-call-expressions).
1580 Type parameters can be specified for a trait to make it generic. These appear
1581 after the trait name, using the same syntax used in [generic
1582 functions](#generic-functions).
1586 fn len(&self) -> u32;
1587 fn elt_at(&self, n: u32) -> T;
1588 fn iter<F>(&self, F) where F: Fn(T);
1592 Generic functions may use traits as _bounds_ on their type parameters. This
1593 will have two effects: only types that have the trait may instantiate the
1594 parameter, and within the generic function, the methods of the trait can be
1595 called on values that have the parameter's type. For example:
1598 # type Surface = i32;
1599 # trait Shape { fn draw(&self, Surface); }
1600 fn draw_twice<T: Shape>(surface: Surface, sh: T) {
1606 Traits also define an [object type](#object-types) with the same name as the
1607 trait. Values of this type are created by [casting](#type-cast-expressions)
1608 pointer values (pointing to a type for which an implementation of the given
1609 trait is in scope) to pointers to the trait name, used as a type.
1613 # impl Shape for i32 { }
1614 # let mycircle = 0i32;
1615 let myshape: Box<Shape> = Box::new(mycircle) as Box<Shape>;
1618 The resulting value is a box containing the value that was cast, along with
1619 information that identifies the methods of the implementation that was used.
1620 Values with a trait type can have [methods called](#method-call-expressions) on
1621 them, for any method in the trait, and can be used to instantiate type
1622 parameters that are bounded by the trait.
1624 Trait methods may be static, which means that they lack a `self` argument.
1625 This means that they can only be called with function call syntax (`f(x)`) and
1626 not method call syntax (`obj.f()`). The way to refer to the name of a static
1627 method is to qualify it with the trait name, treating the trait name like a
1628 module. For example:
1632 fn from_i32(n: i32) -> Self;
1635 fn from_i32(n: i32) -> f64 { n as f64 }
1637 let x: f64 = Num::from_i32(42);
1640 Traits may inherit from other traits. For example, in
1643 trait Shape { fn area() -> f64; }
1644 trait Circle : Shape { fn radius() -> f64; }
1647 the syntax `Circle : Shape` means that types that implement `Circle` must also
1648 have an implementation for `Shape`. Multiple supertraits are separated by `+`,
1649 `trait Circle : Shape + PartialEq { }`. In an implementation of `Circle` for a
1650 given type `T`, methods can refer to `Shape` methods, since the typechecker
1651 checks that any type with an implementation of `Circle` also has an
1652 implementation of `Shape`.
1654 In type-parameterized functions, methods of the supertrait may be called on
1655 values of subtrait-bound type parameters. Referring to the previous example of
1656 `trait Circle : Shape`:
1659 # trait Shape { fn area(&self) -> f64; }
1660 # trait Circle : Shape { fn radius(&self) -> f64; }
1661 fn radius_times_area<T: Circle>(c: T) -> f64 {
1662 // `c` is both a Circle and a Shape
1663 c.radius() * c.area()
1667 Likewise, supertrait methods may also be called on trait objects.
1670 # trait Shape { fn area(&self) -> f64; }
1671 # trait Circle : Shape { fn radius(&self) -> f64; }
1672 # impl Shape for i32 { fn area(&self) -> f64 { 0.0 } }
1673 # impl Circle for i32 { fn radius(&self) -> f64 { 0.0 } }
1674 # let mycircle = 0i32;
1675 let mycircle = Box::new(mycircle) as Box<Circle>;
1676 let nonsense = mycircle.radius() * mycircle.area();
1681 An _implementation_ is an item that implements a [trait](#traits) for a
1684 Implementations are defined with the keyword `impl`.
1687 # struct Point {x: f64, y: f64};
1688 # impl Copy for Point {}
1689 # type Surface = i32;
1690 # struct BoundingBox {x: f64, y: f64, width: f64, height: f64};
1691 # trait Shape { fn draw(&self, Surface); fn bounding_box(&self) -> BoundingBox; }
1692 # fn do_draw_circle(s: Surface, c: Circle) { }
1698 impl Copy for Circle {}
1700 impl Shape for Circle {
1701 fn draw(&self, s: Surface) { do_draw_circle(s, *self); }
1702 fn bounding_box(&self) -> BoundingBox {
1703 let r = self.radius;
1704 BoundingBox{x: self.center.x - r, y: self.center.y - r,
1705 width: 2.0 * r, height: 2.0 * r}
1710 It is possible to define an implementation without referring to a trait. The
1711 methods in such an implementation can only be used as direct calls on the
1712 values of the type that the implementation targets. In such an implementation,
1713 the trait type and `for` after `impl` are omitted. Such implementations are
1714 limited to nominal types (enums, structs), and the implementation must appear
1715 in the same module or a sub-module as the `self` type:
1718 struct Point {x: i32, y: i32}
1722 println!("Point is at ({}, {})", self.x, self.y);
1726 let my_point = Point {x: 10, y:11};
1730 When a trait _is_ specified in an `impl`, all methods declared as part of the
1731 trait must be implemented, with matching types and type parameter counts.
1733 An implementation can take type parameters, which can be different from the
1734 type parameters taken by the trait it implements. Implementation parameters
1735 are written after the `impl` keyword.
1739 impl<T> Seq<T> for Vec<T> {
1742 impl Seq<bool> for u32 {
1743 /* Treat the integer as a sequence of bits */
1750 extern_block_item : "extern" '{' extern_block '}' ;
1751 extern_block : [ foreign_fn ] * ;
1754 External blocks form the basis for Rust's foreign function interface.
1755 Declarations in an external block describe symbols in external, non-Rust
1758 Functions within external blocks are declared in the same way as other Rust
1759 functions, with the exception that they may not have a body and are instead
1760 terminated by a semicolon.
1764 use libc::{c_char, FILE};
1767 fn fopen(filename: *const c_char, mode: *const c_char) -> *mut FILE;
1772 Functions within external blocks may be called by Rust code, just like
1773 functions defined in Rust. The Rust compiler automatically translates between
1774 the Rust ABI and the foreign ABI.
1776 A number of [attributes](#attributes) control the behavior of external blocks.
1778 By default external blocks assume that the library they are calling uses the
1779 standard C "cdecl" ABI. Other ABIs may be specified using an `abi` string, as
1783 // Interface to the Windows API
1784 extern "stdcall" { }
1787 The `link` attribute allows the name of the library to be specified. When
1788 specified the compiler will attempt to link against the native library of the
1792 #[link(name = "crypto")]
1796 The type of a function declared in an extern block is `extern "abi" fn(A1, ...,
1797 An) -> R`, where `A1...An` are the declared types of its arguments and `R` is
1798 the declared return type.
1800 ## Visibility and Privacy
1802 These two terms are often used interchangeably, and what they are attempting to
1803 convey is the answer to the question "Can this item be used at this location?"
1805 Rust's name resolution operates on a global hierarchy of namespaces. Each level
1806 in the hierarchy can be thought of as some item. The items are one of those
1807 mentioned above, but also include external crates. Declaring or defining a new
1808 module can be thought of as inserting a new tree into the hierarchy at the
1809 location of the definition.
1811 To control whether interfaces can be used across modules, Rust checks each use
1812 of an item to see whether it should be allowed or not. This is where privacy
1813 warnings are generated, or otherwise "you used a private item of another module
1814 and weren't allowed to."
1816 By default, everything in Rust is *private*, with one exception. Enum variants
1817 in a `pub` enum are also public by default. You are allowed to alter this
1818 default visibility with the `priv` keyword. When an item is declared as `pub`,
1819 it can be thought of as being accessible to the outside world. For example:
1822 # #![allow(missing_copy_implementations)]
1824 // Declare a private struct
1827 // Declare a public struct with a private field
1832 // Declare a public enum with two public variants
1834 PubliclyAccessibleState,
1835 PubliclyAccessibleState2,
1839 With the notion of an item being either public or private, Rust allows item
1840 accesses in two cases:
1842 1. If an item is public, then it can be used externally through any of its
1844 2. If an item is private, it may be accessed by the current module and its
1847 These two cases are surprisingly powerful for creating module hierarchies
1848 exposing public APIs while hiding internal implementation details. To help
1849 explain, here's a few use cases and what they would entail:
1851 * A library developer needs to expose functionality to crates which link
1852 against their library. As a consequence of the first case, this means that
1853 anything which is usable externally must be `pub` from the root down to the
1854 destination item. Any private item in the chain will disallow external
1857 * A crate needs a global available "helper module" to itself, but it doesn't
1858 want to expose the helper module as a public API. To accomplish this, the
1859 root of the crate's hierarchy would have a private module which then
1860 internally has a "public api". Because the entire crate is a descendant of
1861 the root, then the entire local crate can access this private module through
1864 * When writing unit tests for a module, it's often a common idiom to have an
1865 immediate child of the module to-be-tested named `mod test`. This module
1866 could access any items of the parent module through the second case, meaning
1867 that internal implementation details could also be seamlessly tested from the
1870 In the second case, it mentions that a private item "can be accessed" by the
1871 current module and its descendants, but the exact meaning of accessing an item
1872 depends on what the item is. Accessing a module, for example, would mean
1873 looking inside of it (to import more items). On the other hand, accessing a
1874 function would mean that it is invoked. Additionally, path expressions and
1875 import statements are considered to access an item in the sense that the
1876 import/expression is only valid if the destination is in the current visibility
1879 Here's an example of a program which exemplifies the three cases outlined
1883 // This module is private, meaning that no external crate can access this
1884 // module. Because it is private at the root of this current crate, however, any
1885 // module in the crate may access any publicly visible item in this module.
1886 mod crate_helper_module {
1888 // This function can be used by anything in the current crate
1889 pub fn crate_helper() {}
1891 // This function *cannot* be used by anything else in the crate. It is not
1892 // publicly visible outside of the `crate_helper_module`, so only this
1893 // current module and its descendants may access it.
1894 fn implementation_detail() {}
1897 // This function is "public to the root" meaning that it's available to external
1898 // crates linking against this one.
1899 pub fn public_api() {}
1901 // Similarly to 'public_api', this module is public so external crates may look
1904 use crate_helper_module;
1906 pub fn my_method() {
1907 // Any item in the local crate may invoke the helper module's public
1908 // interface through a combination of the two rules above.
1909 crate_helper_module::crate_helper();
1912 // This function is hidden to any module which is not a descendant of
1914 fn my_implementation() {}
1920 fn test_my_implementation() {
1921 // Because this module is a descendant of `submodule`, it's allowed
1922 // to access private items inside of `submodule` without a privacy
1924 super::my_implementation();
1932 For a rust program to pass the privacy checking pass, all paths must be valid
1933 accesses given the two rules above. This includes all use statements,
1934 expressions, types, etc.
1936 ### Re-exporting and Visibility
1938 Rust allows publicly re-exporting items through a `pub use` directive. Because
1939 this is a public directive, this allows the item to be used in the current
1940 module through the rules above. It essentially allows public access into the
1941 re-exported item. For example, this program is valid:
1944 pub use self::implementation as api;
1946 mod implementation {
1953 This means that any external crate referencing `implementation::f` would
1954 receive a privacy violation, while the path `api::f` would be allowed.
1956 When re-exporting a private item, it can be thought of as allowing the "privacy
1957 chain" being short-circuited through the reexport instead of passing through
1958 the namespace hierarchy as it normally would.
1963 attribute : "#!" ? '[' meta_item ']' ;
1964 meta_item : ident [ '=' literal
1965 | '(' meta_seq ')' ] ? ;
1966 meta_seq : meta_item [ ',' meta_seq ] ? ;
1969 Any item declaration may have an _attribute_ applied to it. Attributes in Rust
1970 are modeled on Attributes in ECMA-335, with the syntax coming from ECMA-334
1971 (C#). An attribute is a general, free-form metadatum that is interpreted
1972 according to name, convention, and language and compiler version. Attributes
1973 may appear as any of:
1975 * A single identifier, the attribute name
1976 * An identifier followed by the equals sign '=' and a literal, providing a
1978 * An identifier followed by a parenthesized list of sub-attribute arguments
1980 Attributes with a bang ("!") after the hash ("#") apply to the item that the
1981 attribute is declared within. Attributes that do not have a bang after the hash
1982 apply to the item that follows the attribute.
1984 An example of attributes:
1987 // General metadata applied to the enclosing module or crate.
1988 #![crate_type = "lib"]
1990 // A function marked as a unit test
1996 // A conditionally-compiled module
1997 #[cfg(target_os="linux")]
2002 // A lint attribute used to suppress a warning/error
2003 #[allow(non_camel_case_types)]
2007 > **Note:** At some point in the future, the compiler will distinguish between
2008 > language-reserved and user-available attributes. Until then, there is
2009 > effectively no difference between an attribute handled by a loadable syntax
2010 > extension and the compiler.
2012 ### Crate-only attributes
2014 - `crate_name` - specify the this crate's crate name.
2015 - `crate_type` - see [linkage](#linkage).
2016 - `feature` - see [compiler features](#compiler-features).
2017 - `no_builtins` - disable optimizing certain code patterns to invocations of
2018 library functions that are assumed to exist
2019 - `no_main` - disable emitting the `main` symbol. Useful when some other
2020 object being linked to defines `main`.
2021 - `no_start` - disable linking to the `native` crate, which specifies the
2022 "start" language item.
2023 - `no_std` - disable linking to the `std` crate.
2025 ### Module-only attributes
2027 - `no_implicit_prelude` - disable injecting `use std::prelude::*` in this
2029 - `path` - specifies the file to load the module from. `#[path="foo.rs"] mod
2030 bar;` is equivalent to `mod bar { /* contents of foo.rs */ }`. The path is
2031 taken relative to the directory that the current module is in.
2033 ### Function-only attributes
2035 - `main` - indicates that this function should be passed to the entry point,
2036 rather than the function in the crate root named `main`.
2037 - `plugin_registrar` - mark this function as the registration point for
2038 [compiler plugins][plugin], such as loadable syntax extensions.
2039 - `start` - indicates that this function should be used as the entry point,
2040 overriding the "start" language item. See the "start" [language
2041 item](#language-items) for more details.
2042 - `test` - indicates that this function is a test function, to only be compiled
2043 in case of `--test`.
2045 ### Static-only attributes
2047 - `thread_local` - on a `static mut`, this signals that the value of this
2048 static may change depending on the current thread. The exact consequences of
2049 this are implementation-defined.
2053 On an `extern` block, the following attributes are interpreted:
2055 - `link_args` - specify arguments to the linker, rather than just the library
2056 name and type. This is feature gated and the exact behavior is
2057 implementation-defined (due to variety of linker invocation syntax).
2058 - `link` - indicate that a native library should be linked to for the
2059 declarations in this block to be linked correctly. `link` supports an optional `kind`
2060 key with three possible values: `dylib`, `static`, and `framework`. See [external blocks](#external-blocks) for more about external blocks. Two
2061 examples: `#[link(name = "readline")]` and
2062 `#[link(name = "CoreFoundation", kind = "framework")]`.
2064 On declarations inside an `extern` block, the following attributes are
2067 - `link_name` - the name of the symbol that this function or static should be
2069 - `linkage` - on a static, this specifies the [linkage
2070 type](http://llvm.org/docs/LangRef.html#linkage-types).
2074 - `repr` - on C-like enums, this sets the underlying type used for
2075 representation. Takes one argument, which is the primitive
2076 type this enum should be represented for, or `C`, which specifies that it
2077 should be the default `enum` size of the C ABI for that platform. Note that
2078 enum representation in C is undefined, and this may be incorrect when the C
2079 code is compiled with certain flags.
2083 - `repr` - specifies the representation to use for this struct. Takes a list
2084 of options. The currently accepted ones are `C` and `packed`, which may be
2085 combined. `C` will use a C ABI compatible struct layout, and `packed` will
2086 remove any padding between fields (note that this is very fragile and may
2087 break platforms which require aligned access).
2089 ### Macro- and plugin-related attributes
2091 - `macro_use` on a `mod` — macros defined in this module will be visible in the
2092 module's parent, after this module has been included.
2094 - `macro_use` on an `extern crate` — load macros from this crate. An optional
2095 list of names `#[macro_use(foo, bar)]` restricts the import to just those
2096 macros named. The `extern crate` must appear at the crate root, not inside
2097 `mod`, which ensures proper function of the [`$crate` macro
2098 variable](book/macros.html#the-variable-$crate).
2100 - `macro_reexport` on an `extern crate` — re-export the named macros.
2102 - `macro_export` - export a macro for cross-crate usage.
2104 - `plugin` on an `extern crate` — load this crate as a [compiler
2105 plugin][plugin]. The `plugin` feature gate is required. Any arguments to
2106 the attribute, e.g. `#[plugin=...]` or `#[plugin(...)]`, are provided to the
2109 - `no_link` on an `extern crate` — even if we load this crate for macros or
2110 compiler plugins, don't link it into the output.
2112 See the [macros section of the
2113 book](book/macros.html#scoping-and-macro-import/export) for more information on
2117 ### Miscellaneous attributes
2119 - `export_name` - on statics and functions, this determines the name of the
2121 - `link_section` - on statics and functions, this specifies the section of the
2122 object file that this item's contents will be placed into.
2123 - `no_mangle` - on any item, do not apply the standard name mangling. Set the
2124 symbol for this item to its identifier.
2125 - `packed` - on structs or enums, eliminate any padding that would be used to
2127 - `simd` - on certain tuple structs, derive the arithmetic operators, which
2128 lower to the target's SIMD instructions, if any; the `simd` feature gate
2129 is necessary to use this attribute.
2130 - `static_assert` - on statics whose type is `bool`, terminates compilation
2131 with an error if it is not initialized to `true`.
2132 - `unsafe_destructor` - allow implementations of the "drop" language item
2133 where the type it is implemented for does not implement the "send" language
2134 item; the `unsafe_destructor` feature gate is needed to use this attribute
2135 - `unsafe_no_drop_flag` - on structs, remove the flag that prevents
2136 destructors from being run twice. Destructors might be run multiple times on
2137 the same object with this attribute.
2138 - `doc` - Doc comments such as `/// foo` are equivalent to `#[doc = "foo"]`.
2139 - `rustc_on_unimplemented` - Write a custom note to be shown along with the error
2140 when the trait is found to be unimplemented on a type.
2141 You may use format arguments like `{T}`, `{A}` to correspond to the
2142 types at the point of use corresponding to the type parameters of the
2143 trait of the same name. `{Self}` will be replaced with the type that is supposed
2144 to implement the trait but doesn't. To use this, the `on_unimplemented` feature gate
2147 ### Conditional compilation
2149 Sometimes one wants to have different compiler outputs from the same code,
2150 depending on build target, such as targeted operating system, or to enable
2153 There are two kinds of configuration options, one that is either defined or not
2154 (`#[cfg(foo)]`), and the other that contains a string that can be checked
2155 against (`#[cfg(bar = "baz")]` (currently only compiler-defined configuration
2156 options can have the latter form).
2159 // The function is only included in the build when compiling for OSX
2160 #[cfg(target_os = "macos")]
2165 // This function is only included when either foo or bar is defined
2166 #[cfg(any(foo, bar))]
2167 fn needs_foo_or_bar() {
2171 // This function is only included when compiling for a unixish OS with a 32-bit
2173 #[cfg(all(unix, target_word_size = "32"))]
2174 fn on_32bit_unix() {
2178 // This function is only included when foo is not defined
2180 fn needs_not_foo() {
2185 This illustrates some conditional compilation can be achieved using the
2186 `#[cfg(...)]` attribute. `any`, `all` and `not` can be used to assemble
2187 arbitrarily complex configurations through nesting.
2189 The following configurations must be defined by the implementation:
2191 * `target_arch = "..."`. Target CPU architecture, such as `"x86"`, `"x86_64"`
2192 `"mips"`, `"powerpc"`, `"arm"`, or `"aarch64"`.
2193 * `target_endian = "..."`. Endianness of the target CPU, either `"little"` or
2195 * `target_family = "..."`. Operating system family of the target, e. g.
2196 `"unix"` or `"windows"`. The value of this configuration option is defined
2197 as a configuration itself, like `unix` or `windows`.
2198 * `target_os = "..."`. Operating system of the target, examples include
2199 `"win32"`, `"macos"`, `"linux"`, `"android"`, `"freebsd"` or `"dragonfly"`.
2200 * `target_word_size = "..."`. Target word size in bits. This is set to `"32"`
2201 for targets with 32-bit pointers, and likewise set to `"64"` for 64-bit
2203 * `unix`. See `target_family`.
2204 * `windows`. See `target_family`.
2206 ### Lint check attributes
2208 A lint check names a potentially undesirable coding pattern, such as
2209 unreachable code or omitted documentation, for the static entity to which the
2212 For any lint check `C`:
2214 * `allow(C)` overrides the check for `C` so that violations will go
2216 * `deny(C)` signals an error after encountering a violation of `C`,
2217 * `forbid(C)` is the same as `deny(C)`, but also forbids changing the lint
2219 * `warn(C)` warns about violations of `C` but continues compilation.
2221 The lint checks supported by the compiler can be found via `rustc -W help`,
2222 along with their default settings. [Compiler
2223 plugins](book/plugin.html#lint-plugins) can provide additional lint checks.
2227 // Missing documentation is ignored here
2228 #[allow(missing_docs)]
2229 pub fn undocumented_one() -> i32 { 1 }
2231 // Missing documentation signals a warning here
2232 #[warn(missing_docs)]
2233 pub fn undocumented_too() -> i32 { 2 }
2235 // Missing documentation signals an error here
2236 #[deny(missing_docs)]
2237 pub fn undocumented_end() -> i32 { 3 }
2241 This example shows how one can use `allow` and `warn` to toggle a particular
2245 #[warn(missing_docs)]
2247 #[allow(missing_docs)]
2249 // Missing documentation is ignored here
2250 pub fn undocumented_one() -> i32 { 1 }
2252 // Missing documentation signals a warning here,
2253 // despite the allow above.
2254 #[warn(missing_docs)]
2255 pub fn undocumented_two() -> i32 { 2 }
2258 // Missing documentation signals a warning here
2259 pub fn undocumented_too() -> i32 { 3 }
2263 This example shows how one can use `forbid` to disallow uses of `allow` for
2267 #[forbid(missing_docs)]
2269 // Attempting to toggle warning signals an error here
2270 #[allow(missing_docs)]
2272 pub fn undocumented_too() -> i32 { 2 }
2278 Some primitive Rust operations are defined in Rust code, rather than being
2279 implemented directly in C or assembly language. The definitions of these
2280 operations have to be easy for the compiler to find. The `lang` attribute
2281 makes it possible to declare these operations. For example, the `str` module
2282 in the Rust standard library defines the string equality function:
2286 pub fn eq_slice(a: &str, b: &str) -> bool {
2291 The name `str_eq` has a special meaning to the Rust compiler, and the presence
2292 of this definition means that it will use this definition when generating calls
2293 to the string equality function.
2295 A complete list of the built-in language items follows:
2297 #### Built-in Traits
2300 : Types that do not move ownership when used by-value.
2304 : Able to be sent across thread boundaries.
2306 : Has a size known at compile time.
2308 : Able to be safely shared between threads when aliased.
2312 These language items are traits:
2315 : Elements can be added (for example, integers and floats).
2317 : Elements can be subtracted.
2319 : Elements can be multiplied.
2321 : Elements have a division operation.
2323 : Elements have a remainder operation.
2325 : Elements can be negated arithmetically.
2327 : Elements can be negated logically.
2329 : Elements have an exclusive-or operation.
2331 : Elements have a bitwise `and` operation.
2333 : Elements have a bitwise `or` operation.
2335 : Elements have a left shift operation.
2337 : Elements have a right shift operation.
2339 : Elements can be indexed.
2341 : ___Needs filling in___
2343 : Elements can be compared for equality.
2345 : Elements have a partial ordering.
2347 : `*` can be applied, yielding a reference to another type.
2349 : `*` can be applied, yielding a mutable reference to another type.
2351 These are functions:
2354 : ___Needs filling in___
2356 : ___Needs filling in___
2358 : ___Needs filling in___
2360 : Compare two strings (`&str`) for equality.
2362 : Return a new unique string
2363 containing a copy of the contents of a unique string.
2368 : The type returned by the `type_id` intrinsic.
2370 : A type whose contents can be mutated through an immutable reference.
2374 These types help drive the compiler's analysis
2377 : ___Needs filling in___
2379 : This type does not implement "copy", even if eligible.
2381 : ___Needs filling in___
2383 : Free memory that was allocated on the exchange heap.
2385 : Allocate memory on the exchange heap.
2386 * `closure_exchange_malloc`
2387 : ___Needs filling in___
2389 : Abort the program with an error.
2390 * `fail_bounds_check`
2391 : Abort the program with a bounds check error.
2393 : Free memory that was allocated on the managed heap.
2395 : ___Needs filling in___
2397 : ___Needs filling in___
2399 : ___Needs filling in___
2400 * `contravariant_lifetime`
2401 : The lifetime parameter should be considered contravariant.
2402 * `covariant_lifetime`
2403 : The lifetime parameter should be considered covariant.
2404 * `invariant_lifetime`
2405 : The lifetime parameter should be considered invariant.
2407 : Allocate memory on the managed heap.
2409 : ___Needs filling in___
2411 : ___Needs filling in___
2413 : ___Needs filling in___
2414 * `contravariant_type`
2415 : The type parameter should be considered contravariant.
2417 : The type parameter should be considered covariant.
2419 : The type parameter should be considered invariant.
2421 : ___Needs filling in___
2423 > **Note:** This list is likely to become out of date. We should auto-generate
2424 > it from `librustc/middle/lang_items.rs`.
2426 ### Inline attributes
2428 The inline attribute is used to suggest to the compiler to perform an inline
2429 expansion and place a copy of the function or static in the caller rather than
2430 generating code to call the function or access the static where it is defined.
2432 The compiler automatically inlines functions based on internal heuristics.
2433 Incorrectly inlining functions can actually making the program slower, so it
2434 should be used with care.
2436 Immutable statics are always considered inlineable unless marked with
2437 `#[inline(never)]`. It is undefined whether two different inlineable statics
2438 have the same memory address. In other words, the compiler is free to collapse
2439 duplicate inlineable statics together.
2441 `#[inline]` and `#[inline(always)]` always causes the function to be serialized
2442 into crate metadata to allow cross-crate inlining.
2444 There are three different types of inline attributes:
2446 * `#[inline]` hints the compiler to perform an inline expansion.
2447 * `#[inline(always)]` asks the compiler to always perform an inline expansion.
2448 * `#[inline(never)]` asks the compiler to never perform an inline expansion.
2452 The `derive` attribute allows certain traits to be automatically implemented
2453 for data structures. For example, the following will create an `impl` for the
2454 `PartialEq` and `Clone` traits for `Foo`, the type parameter `T` will be given
2455 the `PartialEq` or `Clone` constraints for the appropriate `impl`:
2458 #[derive(PartialEq, Clone)]
2465 The generated `impl` for `PartialEq` is equivalent to
2468 # struct Foo<T> { a: i32, b: T }
2469 impl<T: PartialEq> PartialEq for Foo<T> {
2470 fn eq(&self, other: &Foo<T>) -> bool {
2471 self.a == other.a && self.b == other.b
2474 fn ne(&self, other: &Foo<T>) -> bool {
2475 self.a != other.a || self.b != other.b
2480 Supported traits for `derive` are:
2482 * Comparison traits: `PartialEq`, `Eq`, `PartialOrd`, `Ord`.
2483 * Serialization: `Encodable`, `Decodable`. These require `serialize`.
2484 * `Clone`, to create `T` from `&T` via a copy.
2485 * `Default`, to create an empty instance of a data type.
2486 * `FromPrimitive`, to create an instance from a numeric primitive.
2487 * `Hash`, to iterate over the bytes in a data type.
2488 * `Rand`, to create a random instance of a data type.
2489 * `Show`, to format a value using the `{}` formatter.
2490 * `Zero`, to create a zero instance of a numeric data type.
2494 One can indicate the stability of an API using the following attributes:
2496 * `deprecated`: This item should no longer be used, e.g. it has been
2497 replaced. No guarantee of backwards-compatibility.
2498 * `experimental`: This item was only recently introduced or is
2499 otherwise in a state of flux. It may change significantly, or even
2500 be removed. No guarantee of backwards-compatibility.
2501 * `unstable`: This item is still under development, but requires more
2502 testing to be considered stable. No guarantee of backwards-compatibility.
2503 * `stable`: This item is considered stable, and will not change
2504 significantly. Guarantee of backwards-compatibility.
2505 * `frozen`: This item is very stable, and is unlikely to
2506 change. Guarantee of backwards-compatibility.
2507 * `locked`: This item will never change unless a serious bug is
2508 found. Guarantee of backwards-compatibility.
2510 These levels are directly inspired by
2511 [Node.js' "stability index"](http://nodejs.org/api/documentation.html).
2513 Stability levels are inherited, so an item's stability attribute is the default
2514 stability for everything nested underneath it.
2516 There are lints for disallowing items marked with certain levels: `deprecated`,
2517 `experimental` and `unstable`. For now, only `deprecated` warns by default, but
2518 this will change once the standard library has been stabilized. Stability
2519 levels are meant to be promises at the crate level, so these lints only apply
2520 when referencing items from an _external_ crate, not to items defined within
2521 the current crate. Items with no stability level are considered to be unstable
2522 for the purposes of the lint. One can give an optional string that will be
2523 displayed when the lint flags the use of an item.
2525 For example, if we define one crate called `stability_levels`:
2528 #[deprecated="replaced by `best`"]
2530 // delete everything
2534 // delete fewer things
2543 then the lints will work as follows for a client crate:
2547 extern crate stability_levels;
2548 use stability_levels::{bad, better, best};
2551 bad(); // "warning: use of deprecated item: replaced by `best`"
2553 better(); // "warning: use of unmarked item"
2555 best(); // no warning
2559 > **Note:** Currently these are only checked when applied to individual
2560 > functions, structs, methods and enum variants, *not* to entire modules,
2561 > traits, impls or enums themselves.
2563 ### Compiler Features
2565 Certain aspects of Rust may be implemented in the compiler, but they're not
2566 necessarily ready for every-day use. These features are often of "prototype
2567 quality" or "almost production ready", but may not be stable enough to be
2568 considered a full-fledged language feature.
2570 For this reason, Rust recognizes a special crate-level attribute of the form:
2573 #![feature(feature1, feature2, feature3)]
2576 This directive informs the compiler that the feature list: `feature1`,
2577 `feature2`, and `feature3` should all be enabled. This is only recognized at a
2578 crate-level, not at a module-level. Without this directive, all features are
2579 considered off, and using the features will result in a compiler error.
2581 The currently implemented features of the reference compiler are:
2583 * `asm` - The `asm!` macro provides a means for inline assembly. This is often
2584 useful, but the exact syntax for this feature along with its
2585 semantics are likely to change, so this macro usage must be opted
2588 * `concat_idents` - Allows use of the `concat_idents` macro, which is in many
2589 ways insufficient for concatenating identifiers, and may be
2590 removed entirely for something more wholesome.
2592 * `default_type_params` - Allows use of default type parameters. The future of
2593 this feature is uncertain.
2595 * `intrinsics` - Allows use of the "rust-intrinsics" ABI. Compiler intrinsics
2596 are inherently unstable and no promise about them is made.
2598 * `lang_items` - Allows use of the `#[lang]` attribute. Like `intrinsics`,
2599 lang items are inherently unstable and no promise about them
2602 * `link_args` - This attribute is used to specify custom flags to the linker,
2603 but usage is strongly discouraged. The compiler's usage of the
2604 system linker is not guaranteed to continue in the future, and
2605 if the system linker is not used then specifying custom flags
2606 doesn't have much meaning.
2608 * `link_llvm_intrinsics` – Allows linking to LLVM intrinsics via
2609 `#[link_name="llvm.*"]`.
2611 * `linkage` - Allows use of the `linkage` attribute, which is not portable.
2613 * `log_syntax` - Allows use of the `log_syntax` macro attribute, which is a
2614 nasty hack that will certainly be removed.
2616 * `non_ascii_idents` - The compiler supports the use of non-ascii identifiers,
2617 but the implementation is a little rough around the
2618 edges, so this can be seen as an experimental feature
2619 for now until the specification of identifiers is fully
2622 * `once_fns` - Onceness guarantees a closure is only executed once. Defining a
2623 closure as `once` is unlikely to be supported going forward. So
2624 they are hidden behind this feature until they are to be removed.
2626 * `plugin` - Usage of [compiler plugins][plugin] for custom lints or syntax extensions.
2627 These depend on compiler internals and are subject to change.
2629 * `plugin_registrar` - Indicates that a crate provides [compiler plugins][plugin].
2631 * `quote` - Allows use of the `quote_*!` family of macros, which are
2632 implemented very poorly and will likely change significantly
2633 with a proper implementation.
2635 * `rustc_diagnostic_macros`- A mysterious feature, used in the implementation
2636 of rustc, not meant for mortals.
2638 * `simd` - Allows use of the `#[simd]` attribute, which is overly simple and
2639 not the SIMD interface we want to expose in the long term.
2641 * `struct_inherit` - Allows using struct inheritance, which is barely
2642 implemented and will probably be removed. Don't use this.
2644 * `struct_variant` - Structural enum variants (those with named fields). It is
2645 currently unknown whether this style of enum variant is as
2646 fully supported as the tuple-forms, and it's not certain
2647 that this style of variant should remain in the language.
2648 For now this style of variant is hidden behind a feature
2651 * `thread_local` - The usage of the `#[thread_local]` attribute is experimental
2652 and should be seen as unstable. This attribute is used to
2653 declare a `static` as being unique per-thread leveraging
2654 LLVM's implementation which works in concert with the kernel
2655 loader and dynamic linker. This is not necessarily available
2656 on all platforms, and usage of it is discouraged (rust
2657 focuses more on thread-local data instead of thread-local
2660 * `trace_macros` - Allows use of the `trace_macros` macro, which is a nasty
2661 hack that will certainly be removed.
2663 * `unboxed_closures` - Rust's new closure design, which is currently a work in
2664 progress feature with many known bugs.
2666 * `unsafe_destructor` - Allows use of the `#[unsafe_destructor]` attribute,
2667 which is considered wildly unsafe and will be
2668 obsoleted by language improvements.
2670 * `associated_types` - Allows type aliases in traits. Experimental.
2672 If a feature is promoted to a language feature, then all existing programs will
2673 start to receive compilation warnings about #[feature] directives which enabled
2674 the new feature (because the directive is no longer necessary). However, if a
2675 feature is decided to be removed from the language, errors will be issued (if
2676 there isn't a parser error first). The directive in this case is no longer
2677 necessary, and it's likely that existing code will break if the feature isn't
2680 If an unknown feature is found in a directive, it results in a compiler error.
2681 An unknown feature is one which has never been recognized by the compiler.
2683 # Statements and expressions
2685 Rust is _primarily_ an expression language. This means that most forms of
2686 value-producing or effect-causing evaluation are directed by the uniform syntax
2687 category of _expressions_. Each kind of expression can typically _nest_ within
2688 each other kind of expression, and rules for evaluation of expressions involve
2689 specifying both the value produced by the expression and the order in which its
2690 sub-expressions are themselves evaluated.
2692 In contrast, statements in Rust serve _mostly_ to contain and explicitly
2693 sequence expression evaluation.
2697 A _statement_ is a component of a block, which is in turn a component of an
2698 outer [expression](#expressions) or [function](#functions).
2700 Rust has two kinds of statement: [declaration
2701 statements](#declaration-statements) and [expression
2702 statements](#expression-statements).
2704 ### Declaration statements
2706 A _declaration statement_ is one that introduces one or more *names* into the
2707 enclosing statement block. The declared names may denote new slots or new
2710 #### Item declarations
2712 An _item declaration statement_ has a syntactic form identical to an
2713 [item](#items) declaration within a module. Declaring an item — a
2714 function, enumeration, structure, type, static, trait, implementation or module
2715 — locally within a statement block is simply a way of restricting its
2716 scope to a narrow region containing all of its uses; it is otherwise identical
2717 in meaning to declaring the item outside the statement block.
2719 > **Note**: there is no implicit capture of the function's dynamic environment when
2720 > declaring a function-local item.
2722 #### Slot declarations
2725 let_decl : "let" pat [':' type ] ? [ init ] ? ';' ;
2726 init : [ '=' ] expr ;
2729 A _slot declaration_ introduces a new set of slots, given by a pattern. The
2730 pattern may be followed by a type annotation, and/or an initializer expression.
2731 When no type annotation is given, the compiler will infer the type, or signal
2732 an error if insufficient type information is available for definite inference.
2733 Any slots introduced by a slot declaration are visible from the point of
2734 declaration until the end of the enclosing block scope.
2736 ### Expression statements
2738 An _expression statement_ is one that evaluates an [expression](#expressions)
2739 and ignores its result. The type of an expression statement `e;` is always
2740 `()`, regardless of the type of `e`. As a rule, an expression statement's
2741 purpose is to trigger the effects of evaluating its expression.
2745 An expression may have two roles: it always produces a *value*, and it may have
2746 *effects* (otherwise known as "side effects"). An expression *evaluates to* a
2747 value, and has effects during *evaluation*. Many expressions contain
2748 sub-expressions (operands). The meaning of each kind of expression dictates
2751 * Whether or not to evaluate the sub-expressions when evaluating the expression
2752 * The order in which to evaluate the sub-expressions
2753 * How to combine the sub-expressions' values to obtain the value of the expression
2755 In this way, the structure of expressions dictates the structure of execution.
2756 Blocks are just another kind of expression, so blocks, statements, expressions,
2757 and blocks again can recursively nest inside each other to an arbitrary depth.
2759 #### Lvalues, rvalues and temporaries
2761 Expressions are divided into two main categories: _lvalues_ and _rvalues_.
2762 Likewise within each expression, sub-expressions may occur in _lvalue context_
2763 or _rvalue context_. The evaluation of an expression depends both on its own
2764 category and the context it occurs within.
2766 An lvalue is an expression that represents a memory location. These expressions
2767 are [paths](#path-expressions) (which refer to local variables, function and
2768 method arguments, or static variables), dereferences (`*expr`), [indexing
2769 expressions](#index-expressions) (`expr[expr]`), and [field
2770 references](#field-expressions) (`expr.f`). All other expressions are rvalues.
2772 The left operand of an [assignment](#assignment-expressions) or
2773 [compound-assignment](#compound-assignment-expressions) expression is an lvalue
2774 context, as is the single operand of a unary
2775 [borrow](#unary-operator-expressions). All other expression contexts are
2778 When an lvalue is evaluated in an _lvalue context_, it denotes a memory
2779 location; when evaluated in an _rvalue context_, it denotes the value held _in_
2780 that memory location.
2782 When an rvalue is used in an lvalue context, a temporary un-named lvalue is
2783 created and used instead. A temporary's lifetime equals the largest lifetime
2784 of any reference that points to it.
2786 #### Moved and copied types
2788 When a [local variable](#memory-slots) is used as an
2789 [rvalue](#lvalues,-rvalues-and-temporaries) the variable will either be moved
2790 or copied, depending on its type. For types that contain [owning
2791 pointers](#pointer-types) or values that implement the special trait `Drop`,
2792 the variable is moved. All other types are copied.
2794 ### Literal expressions
2796 A _literal expression_ consists of one of the [literal](#literals) forms
2797 described earlier. It directly describes a number, character, string, boolean
2798 value, or the unit value.
2802 "hello"; // string type
2803 '5'; // character type
2807 ### Path expressions
2809 A [path](#paths) used as an expression context denotes either a local variable
2810 or an item. Path expressions are [lvalues](#lvalues,-rvalues-and-temporaries).
2812 ### Tuple expressions
2814 Tuples are written by enclosing zero or more comma-separated expressions in
2815 parentheses. They are used to create [tuple-typed](#tuple-types) values.
2823 ### Unit expressions
2825 The expression `()` denotes the _unit value_, the only value of the type with
2828 ### Structure expressions
2831 struct_expr : expr_path '{' ident ':' expr
2832 [ ',' ident ':' expr ] *
2835 [ ',' expr ] * ')' |
2839 There are several forms of structure expressions. A _structure expression_
2840 consists of the [path](#paths) of a [structure item](#structures), followed by
2841 a brace-enclosed list of one or more comma-separated name-value pairs,
2842 providing the field values of a new instance of the structure. A field name
2843 can be any identifier, and is separated from its value expression by a colon.
2844 The location denoted by a structure field is mutable if and only if the
2845 enclosing structure is mutable.
2847 A _tuple structure expression_ consists of the [path](#paths) of a [structure
2848 item](#structures), followed by a parenthesized list of one or more
2849 comma-separated expressions (in other words, the path of a structure item
2850 followed by a tuple expression). The structure item must be a tuple structure
2853 A _unit-like structure expression_ consists only of the [path](#paths) of a
2854 [structure item](#structures).
2856 The following are examples of structure expressions:
2859 # struct Point { x: f64, y: f64 }
2860 # struct TuplePoint(f64, f64);
2861 # mod game { pub struct User<'a> { pub name: &'a str, pub age: u32, pub score: uint } }
2862 # struct Cookie; fn some_fn<T>(t: T) {}
2863 Point {x: 10.0, y: 20.0};
2864 TuplePoint(10.0, 20.0);
2865 let u = game::User {name: "Joe", age: 35, score: 100_000};
2866 some_fn::<Cookie>(Cookie);
2869 A structure expression forms a new value of the named structure type. Note
2870 that for a given *unit-like* structure type, this will always be the same
2873 A structure expression can terminate with the syntax `..` followed by an
2874 expression to denote a functional update. The expression following `..` (the
2875 base) must have the same structure type as the new structure type being formed.
2876 The entire expression denotes the result of constructing a new structure (with
2877 the same type as the base expression) with the given values for the fields that
2878 were explicitly specified and the values in the base expression for all other
2882 # struct Point3d { x: i32, y: i32, z: i32 }
2883 let base = Point3d {x: 1, y: 2, z: 3};
2884 Point3d {y: 0, z: 10, .. base};
2887 ### Block expressions
2890 block_expr : '{' [ view_item ] *
2891 [ stmt ';' | item ] *
2895 A _block expression_ is similar to a module in terms of the declarations that
2896 are possible. Each block conceptually introduces a new namespace scope. View
2897 items can bring new names into scopes and declared items are in scope for only
2900 A block will execute each statement sequentially, and then execute the
2901 expression (if given). If the final expression is omitted, the type and return
2902 value of the block are `()`, but if it is provided, the type and return value
2903 of the block are that of the expression itself.
2905 ### Method-call expressions
2908 method_call_expr : expr '.' ident paren_expr_list ;
2911 A _method call_ consists of an expression followed by a single dot, an
2912 identifier, and a parenthesized expression-list. Method calls are resolved to
2913 methods on specific traits, either statically dispatching to a method if the
2914 exact `self`-type of the left-hand-side is known, or dynamically dispatching if
2915 the left-hand-side expression is an indirect [object type](#object-types).
2917 ### Field expressions
2920 field_expr : expr '.' ident ;
2923 A _field expression_ consists of an expression followed by a single dot and an
2924 identifier, when not immediately followed by a parenthesized expression-list
2925 (the latter is a [method call expression](#method-call-expressions)). A field
2926 expression denotes a field of a [structure](#structure-types).
2931 (Struct {a: 10, b: 20}).a;
2934 A field access is an [lvalue](#lvalues,-rvalues-and-temporaries) referring to
2935 the value of that field. When the type providing the field inherits mutability,
2936 it can be [assigned](#assignment-expressions) to.
2938 Also, if the type of the expression to the left of the dot is a pointer, it is
2939 automatically dereferenced to make the field access possible.
2941 ### Array expressions
2944 array_expr : '[' "mut" ? vec_elems? ']' ;
2946 array_elems : [expr [',' expr]*] | [expr ';' expr] ;
2949 An [array](#array,-and-slice-types) _expression_ is written by enclosing zero
2950 or more comma-separated expressions of uniform type in square brackets.
2952 In the `[expr ';' expr]` form, the expression after the `';'` must be a
2953 constant expression that can be evaluated at compile time, such as a
2954 [literal](#literals) or a [static item](#static-items).
2958 ["a", "b", "c", "d"];
2959 [0is; 128]; // array with 128 zeros
2960 [0u8, 0u8, 0u8, 0u8];
2963 ### Index expressions
2966 idx_expr : expr '[' expr ']' ;
2969 [Array](#array,-and-slice-types)-typed expressions can be indexed by
2970 writing a square-bracket-enclosed expression (the index) after them. When the
2971 array is mutable, the resulting [lvalue](#lvalues,-rvalues-and-temporaries) can
2974 Indices are zero-based, and may be of any integral type. Vector access is
2975 bounds-checked at run-time. When the check fails, it will put the thread in a
2980 (["a", "b"])[10]; // panics
2983 ### Unary operator expressions
2985 Rust defines three unary operators. They are all written as prefix operators,
2986 before the expression they apply to.
2989 : Negation. May only be applied to numeric types.
2991 : Dereference. When applied to a [pointer](#pointer-types) it denotes the
2992 pointed-to location. For pointers to mutable locations, the resulting
2993 [lvalue](#lvalues,-rvalues-and-temporaries) can be assigned to.
2994 On non-pointer types, it calls the `deref` method of the `std::ops::Deref`
2995 trait, or the `deref_mut` method of the `std::ops::DerefMut` trait (if
2996 implemented by the type and required for an outer expression that will or
2997 could mutate the dereference), and produces the result of dereferencing the
2998 `&` or `&mut` borrowed pointer returned from the overload method.
3001 : Logical negation. On the boolean type, this flips between `true` and
3002 `false`. On integer types, this inverts the individual bits in the
3003 two's complement representation of the value.
3005 ### Binary operator expressions
3008 binop_expr : expr binop expr ;
3011 Binary operators expressions are given in terms of [operator
3012 precedence](#operator-precedence).
3014 #### Arithmetic operators
3016 Binary arithmetic expressions are syntactic sugar for calls to built-in traits,
3017 defined in the `std::ops` module of the `std` library. This means that
3018 arithmetic operators can be overridden for user-defined types. The default
3019 meaning of the operators on standard types is given here.
3022 : Addition and array/string concatenation.
3023 Calls the `add` method on the `std::ops::Add` trait.
3026 Calls the `sub` method on the `std::ops::Sub` trait.
3029 Calls the `mul` method on the `std::ops::Mul` trait.
3032 Calls the `div` method on the `std::ops::Div` trait.
3035 Calls the `rem` method on the `std::ops::Rem` trait.
3037 #### Bitwise operators
3039 Like the [arithmetic operators](#arithmetic-operators), bitwise operators are
3040 syntactic sugar for calls to methods of built-in traits. This means that
3041 bitwise operators can be overridden for user-defined types. The default
3042 meaning of the operators on standard types is given here.
3046 Calls the `bitand` method of the `std::ops::BitAnd` trait.
3049 Calls the `bitor` method of the `std::ops::BitOr` trait.
3052 Calls the `bitxor` method of the `std::ops::BitXor` trait.
3054 : Logical left shift.
3055 Calls the `shl` method of the `std::ops::Shl` trait.
3057 : Logical right shift.
3058 Calls the `shr` method of the `std::ops::Shr` trait.
3060 #### Lazy boolean operators
3062 The operators `||` and `&&` may be applied to operands of boolean type. The
3063 `||` operator denotes logical 'or', and the `&&` operator denotes logical
3064 'and'. They differ from `|` and `&` in that the right-hand operand is only
3065 evaluated when the left-hand operand does not already determine the result of
3066 the expression. That is, `||` only evaluates its right-hand operand when the
3067 left-hand operand evaluates to `false`, and `&&` only when it evaluates to
3070 #### Comparison operators
3072 Comparison operators are, like the [arithmetic
3073 operators](#arithmetic-operators), and [bitwise operators](#bitwise-operators),
3074 syntactic sugar for calls to built-in traits. This means that comparison
3075 operators can be overridden for user-defined types. The default meaning of the
3076 operators on standard types is given here.
3080 Calls the `eq` method on the `std::cmp::PartialEq` trait.
3083 Calls the `ne` method on the `std::cmp::PartialEq` trait.
3086 Calls the `lt` method on the `std::cmp::PartialOrd` trait.
3089 Calls the `gt` method on the `std::cmp::PartialOrd` trait.
3091 : Less than or equal.
3092 Calls the `le` method on the `std::cmp::PartialOrd` trait.
3094 : Greater than or equal.
3095 Calls the `ge` method on the `std::cmp::PartialOrd` trait.
3097 #### Type cast expressions
3099 A type cast expression is denoted with the binary operator `as`.
3101 Executing an `as` expression casts the value on the left-hand side to the type
3102 on the right-hand side.
3104 A numeric value can be cast to any numeric type. A raw pointer value can be
3105 cast to or from any integral type or raw pointer type. Any other cast is
3106 unsupported and will fail to compile.
3108 An example of an `as` expression:
3111 # fn sum(v: &[f64]) -> f64 { 0.0 }
3112 # fn len(v: &[f64]) -> i32 { 0 }
3114 fn avg(v: &[f64]) -> f64 {
3115 let sum: f64 = sum(v);
3116 let sz: f64 = len(v) as f64;
3121 #### Assignment expressions
3123 An _assignment expression_ consists of an
3124 [lvalue](#lvalues,-rvalues-and-temporaries) expression followed by an equals
3125 sign (`=`) and an [rvalue](#lvalues,-rvalues-and-temporaries) expression.
3127 Evaluating an assignment expression [either copies or
3128 moves](#moved-and-copied-types) its right-hand operand to its left-hand
3138 #### Compound assignment expressions
3140 The `+`, `-`, `*`, `/`, `%`, `&`, `|`, `^`, `<<`, and `>>` operators may be
3141 composed with the `=` operator. The expression `lval OP= val` is equivalent to
3142 `lval = lval OP val`. For example, `x = x + 1` may be written as `x += 1`.
3144 Any such expression always has the [`unit`](#primitive-types) type.
3146 #### Operator precedence
3148 The precedence of Rust binary operators is ordered as follows, going from
3151 ```{.text .precedence}
3166 Operators at the same precedence level are evaluated left-to-right. [Unary
3167 operators](#unary-operator-expressions) have the same precedence level and are
3168 stronger than any of the binary operators.
3170 ### Grouped expressions
3172 An expression enclosed in parentheses evaluates to the result of the enclosed
3173 expression. Parentheses can be used to explicitly specify evaluation order
3174 within an expression.
3177 paren_expr : '(' expr ')' ;
3180 An example of a parenthesized expression:
3183 let x: i32 = (2 + 3) * 4;
3187 ### Call expressions
3190 expr_list : [ expr [ ',' expr ]* ] ? ;
3191 paren_expr_list : '(' expr_list ')' ;
3192 call_expr : expr paren_expr_list ;
3195 A _call expression_ invokes a function, providing zero or more input slots and
3196 an optional reference slot to serve as the function's output, bound to the
3197 `lval` on the right hand side of the call. If the function eventually returns,
3198 then the expression completes.
3200 Some examples of call expressions:
3203 # fn add(x: i32, y: i32) -> i32 { 0 }
3205 let x: i32 = add(1i32, 2i32);
3206 let pi: Option<f32> = "3.14".parse();
3209 ### Lambda expressions
3212 ident_list : [ ident [ ',' ident ]* ] ? ;
3213 lambda_expr : '|' ident_list '|' expr ;
3216 A _lambda expression_ (sometimes called an "anonymous function expression")
3217 defines a function and denotes it as a value, in a single expression. A lambda
3218 expression is a pipe-symbol-delimited (`|`) list of identifiers followed by an
3221 A lambda expression denotes a function that maps a list of parameters
3222 (`ident_list`) onto the expression that follows the `ident_list`. The
3223 identifiers in the `ident_list` are the parameters to the function. These
3224 parameters' types need not be specified, as the compiler infers them from
3227 Lambda expressions are most useful when passing functions as arguments to other
3228 functions, as an abbreviation for defining and capturing a separate function.
3230 Significantly, lambda expressions _capture their environment_, which regular
3231 [function definitions](#functions) do not. The exact type of capture depends
3232 on the [function type](#function-types) inferred for the lambda expression. In
3233 the simplest and least-expensive form (analogous to a ```|| { }``` expression),
3234 the lambda expression captures its environment by reference, effectively
3235 borrowing pointers to all outer variables mentioned inside the function.
3236 Alternately, the compiler may infer that a lambda expression should copy or
3237 move values (depending on their type) from the environment into the lambda
3238 expression's captured environment.
3240 In this example, we define a function `ten_times` that takes a higher-order
3241 function argument, and call it with a lambda expression as an argument:
3244 fn ten_times<F>(f: F) where F: Fn(i32) {
3252 ten_times(|j| println!("hello, {}", j));
3258 while_expr : "while" no_struct_literal_expr '{' block '}' ;
3261 A `while` loop begins by evaluating the boolean loop conditional expression.
3262 If the loop conditional expression evaluates to `true`, the loop body block
3263 executes and control returns to the loop conditional expression. If the loop
3264 conditional expression evaluates to `false`, the `while` expression completes.
3279 A `loop` expression denotes an infinite loop.
3282 loop_expr : [ lifetime ':' ] "loop" '{' block '}';
3285 A `loop` expression may optionally have a _label_. If a label is present, then
3286 labeled `break` and `continue` expressions nested within this loop may exit out
3287 of this loop or return control to its head. See [Break
3288 expressions](#break-expressions) and [Continue
3289 expressions](#continue-expressions).
3291 ### Break expressions
3294 break_expr : "break" [ lifetime ];
3297 A `break` expression has an optional _label_. If the label is absent, then
3298 executing a `break` expression immediately terminates the innermost loop
3299 enclosing it. It is only permitted in the body of a loop. If the label is
3300 present, then `break foo` terminates the loop with label `foo`, which need not
3301 be the innermost label enclosing the `break` expression, but must enclose it.
3303 ### Continue expressions
3306 continue_expr : "continue" [ lifetime ];
3309 A `continue` expression has an optional _label_. If the label is absent, then
3310 executing a `continue` expression immediately terminates the current iteration
3311 of the innermost loop enclosing it, returning control to the loop *head*. In
3312 the case of a `while` loop, the head is the conditional expression controlling
3313 the loop. In the case of a `for` loop, the head is the call-expression
3314 controlling the loop. If the label is present, then `continue foo` returns
3315 control to the head of the loop with label `foo`, which need not be the
3316 innermost label enclosing the `break` expression, but must enclose it.
3318 A `continue` expression is only permitted in the body of a loop.
3323 for_expr : "for" pat "in" no_struct_literal_expr '{' block '}' ;
3326 A `for` expression is a syntactic construct for looping over elements provided
3327 by an implementation of `std::iter::Iterator`.
3329 An example of a for loop over the contents of an array:
3333 # fn bar(f: Foo) { }
3338 let v: &[Foo] = &[a, b, c];
3345 An example of a for loop over a series of integers:
3348 # fn bar(b:usize) { }
3349 for i in range(0us, 256) {
3357 if_expr : "if" no_struct_literal_expr '{' block '}'
3360 else_tail : "else" [ if_expr | if_let_expr
3364 An `if` expression is a conditional branch in program control. The form of an
3365 `if` expression is a condition expression, followed by a consequent block, any
3366 number of `else if` conditions and blocks, and an optional trailing `else`
3367 block. The condition expressions must have type `bool`. If a condition
3368 expression evaluates to `true`, the consequent block is executed and any
3369 subsequent `else if` or `else` block is skipped. If a condition expression
3370 evaluates to `false`, the consequent block is skipped and any subsequent `else
3371 if` condition is evaluated. If all `if` and `else if` conditions evaluate to
3372 `false` then any `else` block is executed.
3374 ### Match expressions
3377 match_expr : "match" no_struct_literal_expr '{' match_arm * '}' ;
3379 match_arm : attribute * match_pat "=>" [ expr "," | '{' block '}' ] ;
3381 match_pat : pat [ '|' pat ] * [ "if" expr ] ? ;
3384 A `match` expression branches on a *pattern*. The exact form of matching that
3385 occurs depends on the pattern. Patterns consist of some combination of
3386 literals, destructured arrays or enum constructors, structures and tuples,
3387 variable binding specifications, wildcards (`..`), and placeholders (`_`). A
3388 `match` expression has a *head expression*, which is the value to compare to
3389 the patterns. The type of the patterns must equal the type of the head
3392 In a pattern whose head expression has an `enum` type, a placeholder (`_`)
3393 stands for a *single* data field, whereas a wildcard `..` stands for *all* the
3394 fields of a particular variant. For example:
3397 #![feature(box_syntax)]
3398 enum List<X> { Nil, Cons(X, Box<List<X>>) }
3401 let x: List<i32> = List::Cons(10, box List::Cons(11, box List::Nil));
3404 List::Cons(_, box List::Nil) => panic!("singleton list"),
3405 List::Cons(..) => return,
3406 List::Nil => panic!("empty list")
3411 The first pattern matches lists constructed by applying `Cons` to any head
3412 value, and a tail value of `box Nil`. The second pattern matches _any_ list
3413 constructed with `Cons`, ignoring the values of its arguments. The difference
3414 between `_` and `..` is that the pattern `C(_)` is only type-correct if `C` has
3415 exactly one argument, while the pattern `C(..)` is type-correct for any enum
3416 variant `C`, regardless of how many arguments `C` has.
3418 Used inside an array pattern, `..` stands for any number of elements, when the
3419 `advanced_slice_patterns` feature gate is turned on. This wildcard can be used
3420 at most once for a given array, which implies that it cannot be used to
3421 specifically match elements that are at an unknown distance from both ends of a
3422 array, like `[.., 42, ..]`. If preceded by a variable name, it will bind the
3423 corresponding slice to the variable. Example:
3426 # #![feature(advanced_slice_patterns)]
3427 fn is_symmetric(list: &[u32]) -> bool {
3430 [x, inside.., y] if x == y => is_symmetric(inside),
3436 let sym = &[0, 1, 4, 2, 4, 1, 0];
3437 let not_sym = &[0, 1, 7, 2, 4, 1, 0];
3438 assert!(is_symmetric(sym));
3439 assert!(!is_symmetric(not_sym));
3443 A `match` behaves differently depending on whether or not the head expression
3444 is an [lvalue or an rvalue](#lvalues,-rvalues-and-temporaries). If the head
3445 expression is an rvalue, it is first evaluated into a temporary location, and
3446 the resulting value is sequentially compared to the patterns in the arms until
3447 a match is found. The first arm with a matching pattern is chosen as the branch
3448 target of the `match`, any variables bound by the pattern are assigned to local
3449 variables in the arm's block, and control enters the block.
3451 When the head expression is an lvalue, the match does not allocate a temporary
3452 location (however, a by-value binding may copy or move from the lvalue). When
3453 possible, it is preferable to match on lvalues, as the lifetime of these
3454 matches inherits the lifetime of the lvalue, rather than being restricted to
3455 the inside of the match.
3457 An example of a `match` expression:
3460 #![feature(box_syntax)]
3461 # fn process_pair(a: i32, b: i32) { }
3462 # fn process_ten() { }
3464 enum List<X> { Nil, Cons(X, Box<List<X>>) }
3467 let x: List<i32> = List::Cons(10, box List::Cons(11, box List::Nil));
3470 List::Cons(a, box List::Cons(b, _)) => {
3473 List::Cons(10, _) => {
3486 Patterns that bind variables default to binding to a copy or move of the
3487 matched value (depending on the matched value's type). This can be changed to
3488 bind to a reference by using the `ref` keyword, or to a mutable reference using
3491 Subpatterns can also be bound to variables by the use of the syntax `variable @
3492 subpattern`. For example:
3495 #![feature(box_syntax)]
3497 enum List { Nil, Cons(uint, Box<List>) }
3499 fn is_sorted(list: &List) -> bool {
3501 List::Nil | List::Cons(_, box List::Nil) => true,
3502 List::Cons(x, ref r @ box List::Cons(_, _)) => {
3504 box List::Cons(y, _) => (x <= y) && is_sorted(&**r),
3512 let a = List::Cons(6, box List::Cons(7, box List::Cons(42, box List::Nil)));
3513 assert!(is_sorted(&a));
3518 Patterns can also dereference pointers by using the `&`, `&mut` and `box`
3519 symbols, as appropriate. For example, these two matches on `x: &isize` are
3524 let y = match *x { 0 => "zero", _ => "some" };
3525 let z = match x { &0 => "zero", _ => "some" };
3530 A pattern that's just an identifier, like `Nil` in the previous example, could
3531 either refer to an enum variant that's in scope, or bind a new variable. The
3532 compiler resolves this ambiguity by forbidding variable bindings that occur in
3533 `match` patterns from shadowing names of variants that are in scope. For
3534 example, wherever `List` is in scope, a `match` pattern would not be able to
3535 bind `Nil` as a new name. The compiler interprets a variable pattern `x` as a
3536 binding _only_ if there is no variant named `x` in scope. A convention you can
3537 use to avoid conflicts is simply to name variants with upper-case letters, and
3538 local variables with lower-case letters.
3540 Multiple match patterns may be joined with the `|` operator. A range of values
3541 may be specified with `...`. For example:
3546 let message = match x {
3547 0 | 1 => "not many",
3553 Range patterns only work on scalar types (like integers and characters; not
3554 like arrays and structs, which have sub-components). A range pattern may not
3555 be a sub-range of another range pattern inside the same `match`.
3557 Finally, match patterns can accept *pattern guards* to further refine the
3558 criteria for matching a case. Pattern guards appear after the pattern and
3559 consist of a bool-typed expression following the `if` keyword. A pattern guard
3560 may refer to the variables bound within the pattern they follow.
3563 # let maybe_digit = Some(0);
3564 # fn process_digit(i: i32) { }
3565 # fn process_other(i: i32) { }
3567 let message = match maybe_digit {
3568 Some(x) if x < 10 => process_digit(x),
3569 Some(x) => process_other(x),
3574 ### If let expressions
3577 if_let_expr : "if" "let" pat '=' expr '{' block '}'
3579 else_tail : "else" [ if_expr | if_let_expr | '{' block '}' ] ;
3582 An `if let` expression is semantically identical to an `if` expression but in place
3583 of a condition expression it expects a refutable let statement. If the value of the
3584 expression on the right hand side of the let statement matches the pattern, the corresponding
3585 block will execute, otherwise flow proceeds to the first `else` block that follows.
3590 while_let_expr : "while" "let" pat '=' expr '{' block '}' ;
3593 A `while let` loop is semantically identical to a `while` loop but in place of a
3594 condition expression it expects a refutable let statement. If the value of the
3595 expression on the right hand side of the let statement matches the pattern, the
3596 loop body block executes and control returns to the pattern matching statement.
3597 Otherwise, the while expression completes.
3599 ### Return expressions
3602 return_expr : "return" expr ? ;
3605 Return expressions are denoted with the keyword `return`. Evaluating a `return`
3606 expression moves its argument into the output slot of the current function,
3607 destroys the current function activation frame, and transfers control to the
3610 An example of a `return` expression:
3613 fn max(a: i32, b: i32) -> i32 {
3625 Every slot, item and value in a Rust program has a type. The _type_ of a
3626 *value* defines the interpretation of the memory holding it.
3628 Built-in types and type-constructors are tightly integrated into the language,
3629 in nontrivial ways that are not possible to emulate in user-defined types.
3630 User-defined types have limited capabilities.
3634 The primitive types are the following:
3636 * The "unit" type `()`, having the single "unit" value `()` (occasionally called
3638 * The boolean type `bool` with values `true` and `false`.
3639 * The machine types.
3640 * The machine-dependent integer and floating-point types.
3642 [^unittype]: The "unit" value `()` is *not* a sentinel "null pointer" value for
3643 reference slots; the "unit" type is the implicit return type from functions
3644 otherwise lacking a return type, and can be used in other contexts (such as
3645 message-sending or type-parametric code) as a zero-size type.]
3649 The machine types are the following:
3651 * The unsigned word types `u8`, `u16`, `u32` and `u64`, with values drawn from
3652 the integer intervals [0, 2^8 - 1], [0, 2^16 - 1], [0, 2^32 - 1] and
3653 [0, 2^64 - 1] respectively.
3655 * The signed two's complement word types `i8`, `i16`, `i32` and `i64`, with
3656 values drawn from the integer intervals [-(2^(7)), 2^7 - 1],
3657 [-(2^(15)), 2^15 - 1], [-(2^(31)), 2^31 - 1], [-(2^(63)), 2^63 - 1]
3660 * The IEEE 754-2008 `binary32` and `binary64` floating-point types: `f32` and
3661 `f64`, respectively.
3663 #### Machine-dependent integer types
3665 The `usize` type is an unsigned integer type with the same number of bits as the
3666 platform's pointer type. It can represent every memory address in the process.
3668 The `isize` type is a signed integer type with the same number of bits as the
3669 platform's pointer type. The theoretical upper bound on object and array size
3670 is the maximum `isize` value. This ensures that `isize` can be used to calculate
3671 differences between pointers into an object or array and can address every byte
3672 within an object along with one byte past the end.
3676 The types `char` and `str` hold textual data.
3678 A value of type `char` is a [Unicode scalar value](
3679 http://www.unicode.org/glossary/#unicode_scalar_value) (i.e. a code point that
3680 is not a surrogate), represented as a 32-bit unsigned word in the 0x0000 to
3681 0xD7FF or 0xE000 to 0x10FFFF range. A `[char]` array is effectively an UCS-4 /
3684 A value of type `str` is a Unicode string, represented as an array of 8-bit
3685 unsigned bytes holding a sequence of UTF-8 codepoints. Since `str` is of
3686 unknown size, it is not a _first-class_ type, but can only be instantiated
3687 through a pointer type, such as `&str` or `String`.
3691 A tuple *type* is a heterogeneous product of other types, called the *elements*
3692 of the tuple. It has no nominal name and is instead structurally typed.
3694 Tuple types and values are denoted by listing the types or values of their
3695 elements, respectively, in a parenthesized, comma-separated list.
3697 Because tuple elements don't have a name, they can only be accessed by
3700 The members of a tuple are laid out in memory contiguously, in order specified
3703 An example of a tuple type and its use:
3706 type Pair<'a> = (i32, &'a str);
3707 let p: Pair<'static> = (10, "hello");
3709 assert!(b != "world");
3712 ### Array, and Slice types
3714 Rust has two different types for a list of items:
3716 * `[T; N]`, an 'array'.
3717 * `&[T]`, a 'slice'.
3719 An array has a fixed size, and can be allocated on either the stack or the
3722 A slice is a 'view' into an array. It doesn't own the data it points
3725 An example of each kind:
3728 let vec: Vec<i32> = vec![1, 2, 3];
3729 let arr: [i32; 3] = [1, 2, 3];
3730 let s: &[i32] = vec.as_slice();
3733 As you can see, the `vec!` macro allows you to create a `Vec<T>` easily. The
3734 `vec!` macro is also part of the standard library, rather than the language.
3736 All in-bounds elements of arrays, and slices are always initialized, and access
3737 to an array or slice is always bounds-checked.
3741 A `struct` *type* is a heterogeneous product of other types, called the
3742 *fields* of the type.[^structtype]
3744 [^structtype]: `struct` types are analogous `struct` types in C,
3745 the *record* types of the ML family,
3746 or the *structure* types of the Lisp family.
3748 New instances of a `struct` can be constructed with a [struct
3749 expression](#structure-expressions).
3751 The memory layout of a `struct` is undefined by default to allow for compiler
3752 optimizations like field reordering, but it can be fixed with the
3753 `#[repr(...)]` attribute. In either case, fields may be given in any order in
3754 a corresponding struct *expression*; the resulting `struct` value will always
3755 have the same memory layout.
3757 The fields of a `struct` may be qualified by [visibility
3758 modifiers](#re-exporting-and-visibility), to allow access to data in a
3759 structure outside a module.
3761 A _tuple struct_ type is just like a structure type, except that the fields are
3764 A _unit-like struct_ type is like a structure type, except that it has no
3765 fields. The one value constructed by the associated [structure
3766 expression](#structure-expressions) is the only value that inhabits such a
3769 ### Enumerated types
3771 An *enumerated type* is a nominal, heterogeneous disjoint union type, denoted
3772 by the name of an [`enum` item](#enumerations). [^enumtype]
3774 [^enumtype]: The `enum` type is analogous to a `data` constructor declaration in
3775 ML, or a *pick ADT* in Limbo.
3777 An [`enum` item](#enumerations) declares both the type and a number of *variant
3778 constructors*, each of which is independently named and takes an optional tuple
3781 New instances of an `enum` can be constructed by calling one of the variant
3782 constructors, in a [call expression](#call-expressions).
3784 Any `enum` value consumes as much memory as the largest variant constructor for
3785 its corresponding `enum` type.
3787 Enum types cannot be denoted *structurally* as types, but must be denoted by
3788 named reference to an [`enum` item](#enumerations).
3792 Nominal types — [enumerations](#enumerated-types) and
3793 [structures](#structure-types) — may be recursive. That is, each `enum`
3794 constructor or `struct` field may refer, directly or indirectly, to the
3795 enclosing `enum` or `struct` type itself. Such recursion has restrictions:
3797 * Recursive types must include a nominal type in the recursion
3798 (not mere [type definitions](#type-definitions),
3799 or other structural types such as [arrays](#array,-and-slice-types) or [tuples](#tuple-types)).
3800 * A recursive `enum` item must have at least one non-recursive constructor
3801 (in order to give the recursion a basis case).
3802 * The size of a recursive type must be finite;
3803 in other words the recursive fields of the type must be [pointer types](#pointer-types).
3804 * Recursive type definitions can cross module boundaries, but not module *visibility* boundaries,
3805 or crate boundaries (in order to simplify the module system and type checker).
3807 An example of a *recursive* type and its use:
3812 Cons(T, Box<List<T>>)
3815 let a: List<i32> = List::Cons(7, Box::new(List::Cons(13, Box::new(List::Nil))));
3820 All pointers in Rust are explicit first-class values. They can be copied,
3821 stored into data structures, and returned from functions. There are two
3822 varieties of pointer in Rust:
3825 : These point to memory _owned by some other value_.
3826 A reference type is written `&type` for some lifetime-variable `f`,
3827 or just `&'a type` when you need an explicit lifetime.
3828 Copying a reference is a "shallow" operation:
3829 it involves only copying the pointer itself.
3830 Releasing a reference typically has no effect on the value it points to,
3831 with the exception of temporary values, which are released when the last
3832 reference to them is released.
3834 * Raw pointers (`*`)
3835 : Raw pointers are pointers without safety or liveness guarantees.
3836 Raw pointers are written as `*const T` or `*mut T`,
3837 for example `*const int` means a raw pointer to an integer.
3838 Copying or dropping a raw pointer has no effect on the lifecycle of any
3839 other value. Dereferencing a raw pointer or converting it to any other
3840 pointer type is an [`unsafe` operation](#unsafe-functions).
3841 Raw pointers are generally discouraged in Rust code;
3842 they exist to support interoperability with foreign code,
3843 and writing performance-critical or low-level functions.
3845 The standard library contains additional 'smart pointer' types beyond references
3850 The function type constructor `fn` forms new function types. A function type
3851 consists of a possibly-empty set of function-type modifiers (such as `unsafe`
3852 or `extern`), a sequence of input types and an output type.
3854 An example of a `fn` type:
3857 fn add(x: i32, y: i32) -> i32 {
3861 let mut x = add(5,7);
3863 type Binop = fn(i32, i32) -> i32;
3864 let bo: Binop = add;
3870 ```{.ebnf .notation}
3871 closure_type := [ 'unsafe' ] [ '<' lifetime-list '>' ] '|' arg-list '|'
3872 [ ':' bound-list ] [ '->' type ]
3873 lifetime-list := lifetime | lifetime ',' lifetime-list
3874 arg-list := ident ':' type | ident ':' type ',' arg-list
3875 bound-list := bound | bound '+' bound-list
3876 bound := path | lifetime
3879 The type of a closure mapping an input of type `A` to an output of type `B` is
3880 `|A| -> B`. A closure with no arguments or return values has type `||`.
3882 An example of creating and calling a closure:
3885 let captured_var = 10is;
3887 let closure_no_args = |&:| println!("captured_var={}", captured_var);
3889 let closure_args = |&: arg: isize| -> isize {
3890 println!("captured_var={}, arg={}", captured_var, arg);
3891 arg // Note lack of semicolon after 'arg'
3894 fn call_closure<F: Fn(), G: Fn(isize) -> isize>(c1: F, c2: G) {
3899 call_closure(closure_no_args, closure_args);
3905 Every trait item (see [traits](#traits)) defines a type with the same name as
3906 the trait. This type is called the _object type_ of the trait. Object types
3907 permit "late binding" of methods, dispatched using _virtual method tables_
3908 ("vtables"). Whereas most calls to trait methods are "early bound" (statically
3909 resolved) to specific implementations at compile time, a call to a method on an
3910 object type is only resolved to a vtable entry at compile time. The actual
3911 implementation for each vtable entry can vary on an object-by-object basis.
3913 Given a pointer-typed expression `E` of type `&T` or `Box<T>`, where `T`
3914 implements trait `R`, casting `E` to the corresponding pointer type `&R` or
3915 `Box<R>` results in a value of the _object type_ `R`. This result is
3916 represented as a pair of pointers: the vtable pointer for the `T`
3917 implementation of `R`, and the pointer value of `E`.
3919 An example of an object type:
3923 fn stringify(&self) -> String;
3926 impl Printable for isize {
3927 fn stringify(&self) -> String { self.to_string() }
3930 fn print(a: Box<Printable>) {
3931 println!("{}", a.stringify());
3935 print(Box::new(10is) as Box<Printable>);
3939 In this example, the trait `Printable` occurs as an object type in both the
3940 type signature of `print`, and the cast expression in `main`.
3944 Within the body of an item that has type parameter declarations, the names of
3945 its type parameters are types:
3948 fn map<A: Clone, B: Clone>(f: |A| -> B, xs: &[A]) -> Vec<B> {
3952 let first: B = f(xs[0].clone());
3953 let mut rest: Vec<B> = map(f, xs.slice(1, xs.len()));
3954 rest.insert(0, first);
3959 Here, `first` has type `B`, referring to `map`'s `B` type parameter; and `rest`
3960 has type `Vec<B>`, a vector type with element type `B`.
3964 The special type `self` has a meaning within methods inside an impl item. It
3965 refers to the type of the implicit `self` argument. For example, in:
3969 fn make_string(&self) -> String;
3972 impl Printable for String {
3973 fn make_string(&self) -> String {
3979 `self` refers to the value of type `String` that is the receiver for a call to
3980 the method `make_string`.
3984 Types in Rust are categorized into kinds, based on various properties of the
3985 components of the type. The kinds are:
3988 : Types of this kind can be safely sent between threads.
3989 This kind includes scalars, boxes, procs, and
3990 structural types containing only other owned types.
3991 All `Send` types are `'static`.
3993 : Types of this kind consist of "Plain Old Data"
3994 which can be copied by simply moving bits.
3995 All values of this kind can be implicitly copied.
3996 This kind includes scalars and immutable references,
3997 as well as structural types containing other `Copy` types.
3999 : Types of this kind do not contain any references (except for
4000 references with the `static` lifetime, which are allowed).
4001 This can be a useful guarantee for code
4002 that breaks borrowing assumptions
4003 using [`unsafe` operations](#unsafe-functions).
4005 : This is not strictly a kind,
4006 but its presence interacts with kinds:
4007 the `Drop` trait provides a single method `drop`
4008 that takes no parameters,
4009 and is run when values of the type are dropped.
4010 Such a method is called a "destructor",
4011 and are always executed in "top-down" order:
4012 a value is completely destroyed
4013 before any of the values it owns run their destructors.
4014 Only `Send` types can implement `Drop`.
4017 : Types with destructors, closure environments,
4018 and various other _non-first-class_ types,
4019 are not copyable at all.
4020 Such types can usually only be accessed through pointers,
4021 or in some cases, moved between mutable locations.
4023 Kinds can be supplied as _bounds_ on type parameters, like traits, in which
4024 case the parameter is constrained to types satisfying that kind.
4026 By default, type parameters do not carry any assumed kind-bounds at all. When
4027 instantiating a type parameter, the kind bounds on the parameter are checked to
4028 be the same or narrower than the kind of the type that it is instantiated with.
4030 Sending operations are not part of the Rust language, but are implemented in
4031 the library. Generic functions that send values bound the kind of these values
4034 # Memory and concurrency models
4036 Rust has a memory model centered around concurrently-executing _threads_. Thus
4037 its memory model and its concurrency model are best discussed simultaneously,
4038 as parts of each only make sense when considered from the perspective of the
4041 When reading about the memory model, keep in mind that it is partitioned in
4042 order to support threads; and when reading about threads, keep in mind that their
4043 isolation and communication mechanisms are only possible due to the ownership
4044 and lifetime semantics of the memory model.
4048 A Rust program's memory consists of a static set of *items*, a set of
4049 [threads](#threads) each with its own *stack*, and a *heap*. Immutable portions of
4050 the heap may be shared between threads, mutable portions may not.
4052 Allocations in the stack consist of *slots*, and allocations in the heap
4055 ### Memory allocation and lifetime
4057 The _items_ of a program are those functions, modules and types that have their
4058 value calculated at compile-time and stored uniquely in the memory image of the
4059 rust process. Items are neither dynamically allocated nor freed.
4061 A thread's _stack_ consists of activation frames automatically allocated on entry
4062 to each function as the thread executes. A stack allocation is reclaimed when
4063 control leaves the frame containing it.
4065 The _heap_ is a general term that describes boxes. The lifetime of an
4066 allocation in the heap depends on the lifetime of the box values pointing to
4067 it. Since box values may themselves be passed in and out of frames, or stored
4068 in the heap, heap allocations may outlive the frame they are allocated within.
4070 ### Memory ownership
4072 A thread owns all memory it can *safely* reach through local variables, as well
4073 as boxes and references.
4075 When a thread sends a value that has the `Send` trait to another thread, it loses
4076 ownership of the value sent and can no longer refer to it. This is statically
4077 guaranteed by the combined use of "move semantics", and the compiler-checked
4078 _meaning_ of the `Send` trait: it is only instantiated for (transitively)
4079 sendable kinds of data constructor and pointers, never including references.
4081 When a stack frame is exited, its local allocations are all released, and its
4082 references to boxes are dropped.
4084 When a thread finishes, its stack is necessarily empty and it therefore has no
4085 references to any boxes; the remainder of its heap is immediately freed.
4089 A thread's stack contains slots.
4091 A _slot_ is a component of a stack frame, either a function parameter, a
4092 [temporary](#lvalues,-rvalues-and-temporaries), or a local variable.
4094 A _local variable_ (or *stack-local* allocation) holds a value directly,
4095 allocated within the stack's memory. The value is a part of the stack frame.
4097 Local variables are immutable unless declared otherwise like: `let mut x = ...`.
4099 Function parameters are immutable unless declared with `mut`. The `mut` keyword
4100 applies only to the following parameter (so `|mut x, y|` and `fn f(mut x:
4101 Box<i32>, y: Box<i32>)` declare one mutable variable `x` and one immutable
4104 Methods that take either `self` or `Box<Self>` can optionally place them in a
4105 mutable slot by prefixing them with `mut` (similar to regular arguments):
4109 fn change(mut self) -> Self;
4110 fn modify(mut self: Box<Self>) -> Box<Self>;
4114 Local variables are not initialized when allocated; the entire frame worth of
4115 local variables are allocated at once, on frame-entry, in an uninitialized
4116 state. Subsequent statements within a function may or may not initialize the
4117 local variables. Local variables can be used only after they have been
4118 initialized; this is enforced by the compiler.
4122 A _box_ is a reference to a heap allocation holding another value, which is
4123 constructed by the prefix operator `box`. When the standard library is in use,
4124 the type of a box is `std::owned::Box<T>`.
4126 An example of a box type and value:
4129 let x: Box<i32> = Box::new(10);
4132 Box values exist in 1:1 correspondence with their heap allocation, copying a
4133 box value makes a shallow copy of the pointer. Rust will consider a shallow
4134 copy of a box to move ownership of the value. After a value has been moved,
4135 the source location cannot be used unless it is reinitialized.
4138 let x: Box<i32> = Box::new(10);
4140 // attempting to use `x` will result in an error here
4145 Rust's primary concurrency mechanism is called a **thread**.
4147 ### Communication between threads
4149 Rust threads are isolated and generally unable to interfere with one another's
4150 memory directly, except through [`unsafe` code](#unsafe-functions). All
4151 contact between threads is mediated by safe forms of ownership transfer, and data
4152 races on memory are prohibited by the type system.
4154 When you wish to send data between threads, the values are restricted to the
4155 [`Send` type-kind](#type-kinds). Restricting communication interfaces to this
4156 kind ensures that no references move between threads. Thus access to an entire
4157 data structure can be mediated through its owning "root" value; no further
4158 locking or copying is required to avoid data races within the substructure of
4163 The _lifecycle_ of a threads consists of a finite set of states and events that
4164 cause transitions between the states. The lifecycle states of a thread are:
4171 A thread begins its lifecycle — once it has been spawned — in the
4172 *running* state. In this state it executes the statements of its entry
4173 function, and any functions called by the entry function.
4175 A thread may transition from the *running* state to the *blocked* state any time
4176 it makes a blocking communication call. When the call can be completed —
4177 when a message arrives at a sender, or a buffer opens to receive a message
4178 — then the blocked thread will unblock and transition back to *running*.
4180 A thread may transition to the *panicked* state at any time, due being killed by
4181 some external event or internally, from the evaluation of a `panic!()` macro.
4182 Once *panicking*, a thread unwinds its stack and transitions to the *dead* state.
4183 Unwinding the stack of a thread is done by the thread itself, on its own control
4184 stack. If a value with a destructor is freed during unwinding, the code for the
4185 destructor is run, also on the thread's control stack. Running the destructor
4186 code causes a temporary transition to a *running* state, and allows the
4187 destructor code to cause any subsequent state transitions. The original thread
4188 of unwinding and panicking thereby may suspend temporarily, and may involve
4189 (recursive) unwinding of the stack of a failed destructor. Nonetheless, the
4190 outermost unwinding activity will continue until the stack is unwound and the
4191 thread transitions to the *dead* state. There is no way to "recover" from thread
4192 panics. Once a thread has temporarily suspended its unwinding in the *panicking*
4193 state, a panic occurring from within this destructor results in *hard* panic.
4194 A hard panic currently results in the process aborting.
4196 A thread in the *dead* state cannot transition to other states; it exists only to
4197 have its termination status inspected by other threads, and/or to await
4198 reclamation when the last reference to it drops.
4200 # Runtime services, linkage and debugging
4202 The Rust _runtime_ is a relatively compact collection of Rust code that
4203 provides fundamental services and datatypes to all Rust threads at run-time. It
4204 is smaller and simpler than many modern language runtimes. It is tightly
4205 integrated into the language's execution model of memory, threads, communication
4208 ### Memory allocation
4210 The runtime memory-management system is based on a _service-provider
4211 interface_, through which the runtime requests blocks of memory from its
4212 environment and releases them back to its environment when they are no longer
4213 needed. The default implementation of the service-provider interface consists
4214 of the C runtime functions `malloc` and `free`.
4216 The runtime memory-management system, in turn, supplies Rust threads with
4217 facilities for allocating releasing stacks, as well as allocating and freeing
4222 The runtime provides C and Rust code to assist with various built-in types,
4223 such as arrays, strings, and the low level communication system (ports,
4226 Support for other built-in types such as simple types, tuples and enums is
4227 open-coded by the Rust compiler.
4229 ### Thread scheduling and communication
4231 The runtime provides code to manage inter-thread communication. This includes
4232 the system of thread-lifecycle state transitions depending on the contents of
4233 queues, as well as code to copy values between queues and their recipients and
4234 to serialize values for transmission over operating-system inter-process
4235 communication facilities.
4239 The Rust compiler supports various methods to link crates together both
4240 statically and dynamically. This section will explore the various methods to
4241 link Rust crates together, and more information about native libraries can be
4242 found in the [ffi section of the book][ffi].
4244 In one session of compilation, the compiler can generate multiple artifacts
4245 through the usage of either command line flags or the `crate_type` attribute.
4246 If one or more command line flag is specified, all `crate_type` attributes will
4247 be ignored in favor of only building the artifacts specified by command line.
4249 * `--crate-type=bin`, `#[crate_type = "bin"]` - A runnable executable will be
4250 produced. This requires that there is a `main` function in the crate which
4251 will be run when the program begins executing. This will link in all Rust and
4252 native dependencies, producing a distributable binary.
4254 * `--crate-type=lib`, `#[crate_type = "lib"]` - A Rust library will be produced.
4255 This is an ambiguous concept as to what exactly is produced because a library
4256 can manifest itself in several forms. The purpose of this generic `lib` option
4257 is to generate the "compiler recommended" style of library. The output library
4258 will always be usable by rustc, but the actual type of library may change from
4259 time-to-time. The remaining output types are all different flavors of
4260 libraries, and the `lib` type can be seen as an alias for one of them (but the
4261 actual one is compiler-defined).
4263 * `--crate-type=dylib`, `#[crate_type = "dylib"]` - A dynamic Rust library will
4264 be produced. This is different from the `lib` output type in that this forces
4265 dynamic library generation. The resulting dynamic library can be used as a
4266 dependency for other libraries and/or executables. This output type will
4267 create `*.so` files on linux, `*.dylib` files on osx, and `*.dll` files on
4270 * `--crate-type=staticlib`, `#[crate_type = "staticlib"]` - A static system
4271 library will be produced. This is different from other library outputs in that
4272 the Rust compiler will never attempt to link to `staticlib` outputs. The
4273 purpose of this output type is to create a static library containing all of
4274 the local crate's code along with all upstream dependencies. The static
4275 library is actually a `*.a` archive on linux and osx and a `*.lib` file on
4276 windows. This format is recommended for use in situations such as linking
4277 Rust code into an existing non-Rust application because it will not have
4278 dynamic dependencies on other Rust code.
4280 * `--crate-type=rlib`, `#[crate_type = "rlib"]` - A "Rust library" file will be
4281 produced. This is used as an intermediate artifact and can be thought of as a
4282 "static Rust library". These `rlib` files, unlike `staticlib` files, are
4283 interpreted by the Rust compiler in future linkage. This essentially means
4284 that `rustc` will look for metadata in `rlib` files like it looks for metadata
4285 in dynamic libraries. This form of output is used to produce statically linked
4286 executables as well as `staticlib` outputs.
4288 Note that these outputs are stackable in the sense that if multiple are
4289 specified, then the compiler will produce each form of output at once without
4290 having to recompile. However, this only applies for outputs specified by the
4291 same method. If only `crate_type` attributes are specified, then they will all
4292 be built, but if one or more `--crate-type` command line flag is specified,
4293 then only those outputs will be built.
4295 With all these different kinds of outputs, if crate A depends on crate B, then
4296 the compiler could find B in various different forms throughout the system. The
4297 only forms looked for by the compiler, however, are the `rlib` format and the
4298 dynamic library format. With these two options for a dependent library, the
4299 compiler must at some point make a choice between these two formats. With this
4300 in mind, the compiler follows these rules when determining what format of
4301 dependencies will be used:
4303 1. If a static library is being produced, all upstream dependencies are
4304 required to be available in `rlib` formats. This requirement stems from the
4305 reason that a dynamic library cannot be converted into a static format.
4307 Note that it is impossible to link in native dynamic dependencies to a static
4308 library, and in this case warnings will be printed about all unlinked native
4309 dynamic dependencies.
4311 2. If an `rlib` file is being produced, then there are no restrictions on what
4312 format the upstream dependencies are available in. It is simply required that
4313 all upstream dependencies be available for reading metadata from.
4315 The reason for this is that `rlib` files do not contain any of their upstream
4316 dependencies. It wouldn't be very efficient for all `rlib` files to contain a
4317 copy of `libstd.rlib`!
4319 3. If an executable is being produced and the `-C prefer-dynamic` flag is not
4320 specified, then dependencies are first attempted to be found in the `rlib`
4321 format. If some dependencies are not available in an rlib format, then
4322 dynamic linking is attempted (see below).
4324 4. If a dynamic library or an executable that is being dynamically linked is
4325 being produced, then the compiler will attempt to reconcile the available
4326 dependencies in either the rlib or dylib format to create a final product.
4328 A major goal of the compiler is to ensure that a library never appears more
4329 than once in any artifact. For example, if dynamic libraries B and C were
4330 each statically linked to library A, then a crate could not link to B and C
4331 together because there would be two copies of A. The compiler allows mixing
4332 the rlib and dylib formats, but this restriction must be satisfied.
4334 The compiler currently implements no method of hinting what format a library
4335 should be linked with. When dynamically linking, the compiler will attempt to
4336 maximize dynamic dependencies while still allowing some dependencies to be
4337 linked in via an rlib.
4339 For most situations, having all libraries available as a dylib is recommended
4340 if dynamically linking. For other situations, the compiler will emit a
4341 warning if it is unable to determine which formats to link each library with.
4343 In general, `--crate-type=bin` or `--crate-type=lib` should be sufficient for
4344 all compilation needs, and the other options are just available if more
4345 fine-grained control is desired over the output format of a Rust crate.
4347 # Appendix: Rationales and design tradeoffs
4351 # Appendix: Influences
4353 Rust is not a particularly original language, with design elements coming from
4354 a wide range of sources. Some of these are listed below (including elements
4355 that have since been removed):
4357 * SML, OCaml: algebraic datatypes, pattern matching, type inference,
4358 semicolon statement separation
4359 * C++: references, RAII, smart pointers, move semantics, monomorphisation,
4361 * ML Kit, Cyclone: region based memory management
4362 * Haskell (GHC): typeclasses, type families
4363 * Newsqueak, Alef, Limbo: channels, concurrency
4364 * Erlang: message passing, task failure, ~~linked task failure~~,
4365 ~~lightweight concurrency~~
4366 * Swift: optional bindings
4367 * Scheme: hygienic macros
4369 * Ruby: ~~block syntax~~
4370 * NIL, Hermes: ~~typestate~~
4371 * [Unicode Annex #31](http://www.unicode.org/reports/tr31/): identifier and
4374 [ffi]: book/ffi.html
4375 [plugin]: book/plugin.html