1 % The Rust Reference Manual
5 This document is the reference manual 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 a tutorial introduction to the
17 language. Background familiarity with the language is assumed. A separate
18 [tutorial] document is available to help acquire such background familiarity.
20 This document also does not serve as a reference to the [standard] or [extra]
21 libraries included in the language distribution. Those libraries are
22 documented separately by extracting documentation attributes from their
25 [tutorial]: tutorial.html
26 [standard]: std/index.html
27 [extra]: extra/index.html
31 Rust is a work in progress. The language continues to evolve as the design
32 shifts and is fleshed out in working code. Certain parts work, certain parts
33 do not, certain parts will be removed or changed.
35 This manual is a snapshot written in the present tense. All features described
36 exist in working code unless otherwise noted, but some are quite primitive or
37 remain to be further modified by planned work. Some may be temporary. It is a
38 *draft*, and we ask that you not take anything you read here as final.
40 If you have suggestions to make, please try to focus them on *reductions* to
41 the language: possible features that can be combined or omitted. We aim to
42 keep the size and complexity of the language under control.
44 > **Note:** The grammar for Rust given in this document is rough and
45 > very incomplete; only a modest number of sections have accompanying grammar
46 > rules. Formalizing the grammar accepted by the Rust parser is ongoing work,
47 > but future versions of this document will contain a complete
48 > grammar. Moreover, we hope that this grammar will be extracted and verified
49 > as LL(1) by an automated grammar-analysis tool, and further tested against the
50 > Rust sources. Preliminary versions of this automation exist, but are not yet
55 Rust's grammar is defined over Unicode codepoints, each conventionally
56 denoted `U+XXXX`, for 4 or more hexadecimal digits `X`. _Most_ of Rust's
57 grammar is confined to the ASCII range of Unicode, and is described in this
58 document by a dialect of Extended Backus-Naur Form (EBNF), specifically a
59 dialect of EBNF supported by common automated LL(k) parsing tools such as
60 `llgen`, rather than the dialect given in ISO 14977. The dialect can be
61 defined self-referentially as follows:
63 ~~~~ {.notrust .ebnf .notation}
65 rule : nonterminal ':' productionrule ';' ;
66 productionrule : production [ '|' production ] * ;
68 term : element repeats ;
69 element : LITERAL | IDENTIFIER | '[' productionrule ']' ;
70 repeats : [ '*' | '+' ] NUMBER ? | NUMBER ? | '?' ;
75 - Whitespace in the grammar is ignored.
76 - Square brackets are used to group rules.
77 - `LITERAL` is a single printable ASCII character, or an escaped hexadecimal
78 ASCII code of the form `\xQQ`, in single quotes, denoting the corresponding
79 Unicode codepoint `U+00QQ`.
80 - `IDENTIFIER` is a nonempty string of ASCII letters and underscores.
81 - The `repeat` forms apply to the adjacent `element`, and are as follows:
82 - `?` means zero or one repetition
83 - `*` means zero or more repetitions
84 - `+` means one or more repetitions
85 - NUMBER trailing a repeat symbol gives a maximum repetition count
86 - NUMBER on its own gives an exact repetition count
88 This EBNF dialect should hopefully be familiar to many readers.
90 ## Unicode productions
92 A few productions in Rust's grammar permit Unicode codepoints outside the ASCII range.
93 We define these productions in terms of character properties specified in the Unicode standard,
94 rather than in terms of ASCII-range codepoints.
95 The section [Special Unicode Productions](#special-unicode-productions) lists these productions.
97 ## String table productions
99 Some rules in the grammar -- notably [unary
100 operators](#unary-operator-expressions), [binary
101 operators](#binary-operator-expressions), and [keywords](#keywords) --
102 are given in a simplified form: as a listing of a table of unquoted,
103 printable whitespace-separated strings. These cases form a subset of
104 the rules regarding the [token](#tokens) rule, and are assumed to be
105 the result of a lexical-analysis phase feeding the parser, driven by a
106 DFA, operating over the disjunction of all such string table entries.
108 When such a string enclosed in double-quotes (`"`) occurs inside the
109 grammar, it is an implicit reference to a single member of such a string table
110 production. See [tokens](#tokens) for more information.
116 Rust input is interpreted as a sequence of Unicode codepoints encoded in UTF-8,
117 normalized to Unicode normalization form NFKC.
118 Most Rust grammar rules are defined in terms of printable ASCII-range codepoints,
119 but a small number are defined in terms of Unicode properties or explicit codepoint lists.
120 ^[Substitute definitions for the special Unicode productions are provided to the grammar verifier, restricted to ASCII range, when verifying the grammar in this document.]
122 ## Special Unicode Productions
124 The following productions in the Rust grammar are defined in terms of Unicode properties:
125 `ident`, `non_null`, `non_star`, `non_eol`, `non_slash_or_star`, `non_single_quote` and `non_double_quote`.
129 The `ident` production is any nonempty Unicode string of the following form:
131 - The first character has property `XID_start`
132 - The remaining characters have property `XID_continue`
134 that does _not_ occur in the set of [keywords](#keywords).
136 Note: `XID_start` and `XID_continue` as character properties cover the
137 character ranges used to form the more familiar C and Java language-family
140 ### Delimiter-restricted productions
142 Some productions are defined by exclusion of particular Unicode characters:
144 - `non_null` is any single Unicode character aside from `U+0000` (null)
145 - `non_eol` is `non_null` restricted to exclude `U+000A` (`'\n'`)
146 - `non_star` is `non_null` restricted to exclude `U+002A` (`*`)
147 - `non_slash_or_star` is `non_null` restricted to exclude `U+002F` (`/`) and `U+002A` (`*`)
148 - `non_single_quote` is `non_null` restricted to exclude `U+0027` (`'`)
149 - `non_double_quote` is `non_null` restricted to exclude `U+0022` (`"`)
153 ~~~~ {.notrust .ebnf .gram}
154 comment : block_comment | line_comment ;
155 block_comment : "/*" block_comment_body * '*' + '/' ;
156 block_comment_body : [block_comment | character] * ;
157 line_comment : "//" non_eol * ;
160 Comments in Rust code follow the general C++ style of line and block-comment forms,
161 with no nesting of block-comment delimiters.
163 Line comments beginning with exactly _three_ slashes (`///`), and block
164 comments beginning with a exactly one repeated asterisk in the block-open
165 sequence (`/**`), are interpreted as a special syntax for `doc`
166 [attributes](#attributes). That is, they are equivalent to writing
167 `#[doc="..."]` around the body of the comment (this includes the comment
168 characters themselves, ie `/// Foo` turns into `#[doc="/// Foo"]`).
170 Non-doc comments are interpreted as a form of whitespace.
174 ~~~~ {.notrust .ebnf .gram}
175 whitespace_char : '\x20' | '\x09' | '\x0a' | '\x0d' ;
176 whitespace : [ whitespace_char | comment ] + ;
179 The `whitespace_char` production is any nonempty Unicode string consisting of any
180 of the following Unicode characters: `U+0020` (space, `' '`), `U+0009` (tab,
181 `'\t'`), `U+000A` (LF, `'\n'`), `U+000D` (CR, `'\r'`).
183 Rust is a "free-form" language, meaning that all forms of whitespace serve
184 only to separate _tokens_ in the grammar, and have no semantic significance.
186 A Rust program has identical meaning if each whitespace element is replaced
187 with any other legal whitespace element, such as a single space character.
191 ~~~~ {.notrust .ebnf .gram}
192 simple_token : keyword | unop | binop ;
193 token : simple_token | ident | literal | symbol | whitespace token ;
196 Tokens are primitive productions in the grammar defined by regular
197 (non-recursive) languages. "Simple" tokens are given in [string table
198 production](#string-table-productions) form, and occur in the rest of the
199 grammar as double-quoted strings. Other tokens have exact rules given.
203 The keywords are the following strings:
205 ~~~~ {.notrust .keyword}
216 self static struct super
222 Each of these keywords has special meaning in its grammar,
223 and all of them are excluded from the `ident` rule.
227 A literal is an expression consisting of a single token, rather than a
228 sequence of tokens, that immediately and directly denotes the value it
229 evaluates to, rather than referring to it by name or some other evaluation
230 rule. A literal is a form of constant expression, so is evaluated (primarily)
233 ~~~~ {.notrust .ebnf .gram}
234 literal : string_lit | char_lit | num_lit ;
237 #### Character and string literals
239 ~~~~ {.notrust .ebnf .gram}
240 char_lit : '\x27' char_body '\x27' ;
241 string_lit : '"' string_body * '"' | 'r' raw_string ;
243 char_body : non_single_quote
244 | '\x5c' [ '\x27' | common_escape ] ;
246 string_body : non_double_quote
247 | '\x5c' [ '\x22' | common_escape ] ;
248 raw_string : '"' raw_string_body '"' | '#' raw_string '#' ;
250 common_escape : '\x5c'
251 | 'n' | 'r' | 't' | '0'
256 hex_digit : 'a' | 'b' | 'c' | 'd' | 'e' | 'f'
257 | 'A' | 'B' | 'C' | 'D' | 'E' | 'F'
259 oct_digit : '0' | '1' | '2' | '3' | '4' | '5' | '6' | '7' ;
260 dec_digit : '0' | nonzero_dec ;
261 nonzero_dec: '1' | '2' | '3' | '4'
262 | '5' | '6' | '7' | '8' | '9' ;
265 A _character literal_ is a single Unicode character enclosed within two
266 `U+0027` (single-quote) characters, with the exception of `U+0027` itself,
267 which must be _escaped_ by a preceding U+005C character (`\`).
269 A _string literal_ is a sequence of any Unicode characters enclosed within
270 two `U+0022` (double-quote) characters, with the exception of `U+0022`
271 itself, which must be _escaped_ by a preceding `U+005C` character (`\`),
272 or a _raw string literal_.
274 Some additional _escapes_ are available in either character or non-raw string
275 literals. An escape starts with a `U+005C` (`\`) and continues with one of
278 * An _8-bit codepoint escape_ escape starts with `U+0078` (`x`) and is
279 followed by exactly two _hex digits_. It denotes the Unicode codepoint
280 equal to the provided hex value.
281 * A _16-bit codepoint escape_ starts with `U+0075` (`u`) and is followed
282 by exactly four _hex digits_. It denotes the Unicode codepoint equal to
283 the provided hex value.
284 * A _32-bit codepoint escape_ starts with `U+0055` (`U`) and is followed
285 by exactly eight _hex digits_. It denotes the Unicode codepoint equal to
286 the provided hex value.
287 * A _whitespace escape_ is one of the characters `U+006E` (`n`), `U+0072`
288 (`r`), or `U+0074` (`t`), denoting the unicode values `U+000A` (LF),
289 `U+000D` (CR) or `U+0009` (HT) respectively.
290 * The _backslash escape_ is the character `U+005C` (`\`) which must be
291 escaped in order to denote *itself*.
293 Raw string literals do not process any escapes. They start with the character
294 `U+0072` (`r`), followed zero or more of the character `U+0023` (`#`) and a
295 `U+0022` (double-quote) character. The _raw string body_ is not defined in the
296 EBNF grammar above: it can contain any sequence of Unicode characters and is
297 terminated only by another `U+0022` (double-quote) character, followed by the
298 same number of `U+0023` (`#`) characters that preceeded the opening `U+0022`
299 (double-quote) character.
301 All Unicode characters contained in the raw string body represent themselves,
302 the characters `U+0022` (double-quote) (except when followed by at least as
303 many `U+0023` (`#`) characters as were used to start the raw string literal) or
304 `U+005C` (`\`) do not have any special meaning.
306 Examples for string literals:
309 "foo"; r"foo"; // foo
310 "\"foo\""; r#""foo""#; // "foo"
313 r##"foo #"# bar"##; // foo #"# bar
315 "\x52"; "R"; r"R"; // R
316 "\\x52"; r"\x52"; // \x52
321 ~~~~ {.notrust .ebnf .gram}
322 num_lit : nonzero_dec [ dec_digit | '_' ] * num_suffix ?
323 | '0' [ [ dec_digit | '_' ] * num_suffix ?
324 | 'b' [ '1' | '0' | '_' ] + int_suffix ?
325 | 'o' [ oct_digit | '_' ] + int_suffix ?
326 | 'x' [ hex_digit | '_' ] + int_suffix ? ] ;
328 num_suffix : int_suffix | float_suffix ;
330 int_suffix : 'u' int_suffix_size ?
331 | 'i' int_suffix_size ? ;
332 int_suffix_size : [ '8' | '1' '6' | '3' '2' | '6' '4' ] ;
334 float_suffix : [ exponent | '.' dec_lit exponent ? ] ? float_suffix_ty ? ;
335 float_suffix_ty : 'f' [ '3' '2' | '6' '4' ] ;
336 exponent : ['E' | 'e'] ['-' | '+' ] ? dec_lit ;
337 dec_lit : [ dec_digit | '_' ] + ;
340 A _number literal_ is either an _integer literal_ or a _floating-point
341 literal_. The grammar for recognizing the two kinds of literals is mixed,
342 as they are differentiated by suffixes.
344 ##### Integer literals
346 An _integer literal_ has one of four forms:
348 * A _decimal literal_ starts with a *decimal digit* and continues with any
349 mixture of *decimal digits* and _underscores_.
350 * A _hex literal_ starts with the character sequence `U+0030` `U+0078`
351 (`0x`) and continues as any mixture hex digits and underscores.
352 * An _octal literal_ starts with the character sequence `U+0030` `U+006F`
353 (`0o`) and continues as any mixture octal digits and underscores.
354 * A _binary literal_ starts with the character sequence `U+0030` `U+0062`
355 (`0b`) and continues as any mixture binary digits and underscores.
357 An integer literal may be followed (immediately, without any spaces) by an
358 _integer suffix_, which changes the type of the literal. There are two kinds
359 of integer literal suffix:
361 * The `i` and `u` suffixes give the literal type `int` or `uint`,
363 * Each of the signed and unsigned machine types `u8`, `i8`,
364 `u16`, `i16`, `u32`, `i32`, `u64` and `i64`
365 give the literal the corresponding machine type.
367 The type of an _unsuffixed_ integer literal is determined by type inference.
368 If a integer type can be _uniquely_ determined from the surrounding program
369 context, the unsuffixed integer literal has that type. If the program context
370 underconstrains the type, the unsuffixed integer literal's type is `int`; if
371 the program context overconstrains the type, it is considered a static type
374 Examples of integer literals of various forms:
377 123; 0xff00; // type determined by program context
378 // defaults to int in absence of type
384 0o70_i16; // type i16
385 0b1111_1111_1001_0000_i32; // type i32
388 ##### Floating-point literals
390 A _floating-point literal_ has one of two forms:
392 * Two _decimal literals_ separated by a period
393 character `U+002E` (`.`), with an optional _exponent_ trailing after the
394 second decimal literal.
395 * A single _decimal literal_ followed by an _exponent_.
397 By default, a floating-point literal has a generic type, but will fall back to
398 `f64`. A floating-point literal may be followed (immediately, without any
399 spaces) by a _floating-point suffix_, which changes the type of the literal.
400 There are two floating-point suffixes: `f32`, and `f64` (the 32-bit and 64-bit
401 floating point types).
403 Examples of floating-point literals of various forms:
409 12E+99_f64; // type f64
412 ##### Unit and boolean literals
414 The _unit value_, the only value of the type that has the same name, is written as `()`.
415 The two values of the boolean type are written `true` and `false`.
419 ~~~~ {.notrust .ebnf .gram}
421 | '#' | '[' | ']' | '(' | ')' | '{' | '}'
425 Symbols are a general class of printable [token](#tokens) that play structural
426 roles in a variety of grammar productions. They are catalogued here for
427 completeness as the set of remaining miscellaneous printable tokens that do not
428 otherwise appear as [unary operators](#unary-operator-expressions), [binary
429 operators](#binary-operator-expressions), or [keywords](#keywords).
434 ~~~~ {.notrust .ebnf .gram}
435 expr_path : ident [ "::" expr_path_tail ] + ;
436 expr_path_tail : '<' type_expr [ ',' type_expr ] + '>'
439 type_path : ident [ type_path_tail ] + ;
440 type_path_tail : '<' type_expr [ ',' type_expr ] + '>'
444 A _path_ is a sequence of one or more path components _logically_ separated by
445 a namespace qualifier (`::`). If a path consists of only one component, it may
446 refer to either an [item](#items) or a [slot](#memory-slots) in a local
447 control scope. If a path has multiple components, it refers to an item.
449 Every item has a _canonical path_ within its crate, but the path naming an
450 item is only meaningful within a given crate. There is no global namespace
451 across crates; an item's canonical path merely identifies it within the crate.
453 Two examples of simple paths consisting of only identifier components:
460 Path components are usually [identifiers](#identifiers), but the trailing
461 component of a path may be an angle-bracket-enclosed list of type
462 arguments. In [expression](#expressions) context, the type argument list is
463 given after a final (`::`) namespace qualifier in order to disambiguate it
464 from a relational expression involving the less-than symbol (`<`). In type
465 expression context, the final namespace qualifier is omitted.
467 Two examples of paths with type arguments:
470 # struct HashMap<K, V>;
472 # fn id<T>(t: T) -> T { t }
473 type T = HashMap<int,~str>; // Type arguments used in a type expression
474 let x = id::<int>(10); // Type arguments used in a call expression
480 A number of minor features of Rust are not central enough to have their own
481 syntax, and yet are not implementable as functions. Instead, they are given
482 names, and invoked through a consistent syntax: `name!(...)`. Examples
485 * `format!` : format data into a string
486 * `env!` : look up an environment variable's value at compile time
487 * `file!`: return the path to the file being compiled
488 * `stringify!` : pretty-print the Rust expression given as an argument
489 * `include!` : include the Rust expression in the given file
490 * `include_str!` : include the contents of the given file as a string
491 * `include_bin!` : include the contents of the given file as a binary blob
492 * `error!`, `warn!`, `info!`, `debug!` : provide diagnostic information.
494 All of the above extensions are expressions with values.
498 ~~~~ {.notrust .ebnf .gram}
499 expr_macro_rules : "macro_rules" '!' ident '(' macro_rule * ')' ;
500 macro_rule : '(' matcher * ')' "=>" '(' transcriber * ')' ';' ;
501 matcher : '(' matcher * ')' | '[' matcher * ']'
502 | '{' matcher * '}' | '$' ident ':' ident
503 | '$' '(' matcher * ')' sep_token? [ '*' | '+' ]
504 | non_special_token ;
505 transcriber : '(' transcriber * ')' | '[' transcriber * ']'
506 | '{' transcriber * '}' | '$' ident
507 | '$' '(' transcriber * ')' sep_token? [ '*' | '+' ]
508 | non_special_token ;
511 User-defined syntax extensions are called "macros",
512 and the `macro_rules` syntax extension defines them.
513 Currently, user-defined macros can expand to expressions, statements, or items.
515 (A `sep_token` is any token other than `*` and `+`.
516 A `non_special_token` is any token other than a delimiter or `$`.)
518 The macro expander looks up macro invocations by name,
519 and tries each macro rule in turn.
520 It transcribes the first successful match.
521 Matching and transcription are closely related to each other,
522 and we will describe them together.
526 The macro expander matches and transcribes every token that does not begin with a `$` literally, including delimiters.
527 For parsing reasons, delimiters must be balanced, but they are otherwise not special.
529 In the matcher, `$` _name_ `:` _designator_ matches the nonterminal in the
530 Rust syntax named by _designator_. Valid designators are `item`, `block`,
531 `stmt`, `pat`, `expr`, `ty` (type), `ident`, `path`, `matchers` (lhs of the `=>` in macro rules),
532 `tt` (rhs of the `=>` in macro rules). In the transcriber, the designator is already known, and so only
533 the name of a matched nonterminal comes after the dollar sign.
535 In both the matcher and transcriber, the Kleene star-like operator indicates repetition.
536 The Kleene star operator consists of `$` and parens, optionally followed by a separator token, followed by `*` or `+`.
537 `*` means zero or more repetitions, `+` means at least one repetition.
538 The parens are not matched or transcribed.
539 On the matcher side, a name is bound to _all_ of the names it
540 matches, in a structure that mimics the structure of the repetition
541 encountered on a successful match. The job of the transcriber is to sort that
544 The rules for transcription of these repetitions are called "Macro By Example".
545 Essentially, one "layer" of repetition is discharged at a time, and all of
546 them must be discharged by the time a name is transcribed. Therefore,
547 `( $( $i:ident ),* ) => ( $i )` is an invalid macro, but
548 `( $( $i:ident ),* ) => ( $( $i:ident ),* )` is acceptable (if trivial).
550 When Macro By Example encounters a repetition, it examines all of the `$`
551 _name_ s that occur in its body. At the "current layer", they all must repeat
552 the same number of times, so
553 ` ( $( $i:ident ),* ; $( $j:ident ),* ) => ( $( ($i,$j) ),* )` is valid if
554 given the argument `(a,b,c ; d,e,f)`, but not `(a,b,c ; d,e)`. The repetition
555 walks through the choices at that layer in lockstep, so the former input
556 transcribes to `( (a,d), (b,e), (c,f) )`.
558 Nested repetitions are allowed.
560 ### Parsing limitations
562 The parser used by the macro system is reasonably powerful, but the parsing of
563 Rust syntax is restricted in two ways:
565 1. The parser will always parse as much as possible. If it attempts to match
566 `$i:expr [ , ]` against `8 [ , ]`, it will attempt to parse `i` as an array
567 index operation and fail. Adding a separator can solve this problem.
568 2. The parser must have eliminated all ambiguity by the time it reaches a `$` _name_ `:` _designator_.
569 This requirement most often affects name-designator pairs when they occur at the beginning of, or immediately after, a `$(...)*`; requiring a distinctive token in front can solve the problem.
571 ## Syntax extensions useful for the macro author
573 * `log_syntax!` : print out the arguments at compile time
574 * `trace_macros!` : supply `true` or `false` to enable or disable macro expansion logging
575 * `stringify!` : turn the identifier argument into a string literal
576 * `concat!` : concatenates a comma-separated list of literals
577 * `concat_idents!` : create a new identifier by concatenating the arguments
579 # Crates and source files
581 Rust is a *compiled* language.
582 Its semantics obey a *phase distinction* between compile-time and run-time.
583 Those semantic rules that have a *static interpretation* govern the success or failure of compilation.
584 We refer to these rules as "static semantics".
585 Semantic rules called "dynamic semantics" govern the behavior of programs at run-time.
586 A program that fails to compile due to violation of a compile-time rule has no defined dynamic semantics; the compiler should halt with an error report, and produce no executable artifact.
588 The compilation model centres on artifacts called _crates_.
589 Each compilation processes a single crate in source form, and if successful, produces a single crate in binary form: either an executable or a library.^[A crate is somewhat
590 analogous to an *assembly* in the ECMA-335 CLI model, a *library* in the
591 SML/NJ Compilation Manager, a *unit* in the Owens and Flatt module system,
592 or a *configuration* in Mesa.]
594 A _crate_ is a unit of compilation and linking, as well as versioning, distribution and runtime loading.
595 A crate contains a _tree_ of nested [module](#modules) scopes.
596 The top level of this tree is a module that is anonymous (from the point of view of paths within the module) and any item within a crate has a canonical [module path](#paths) denoting its location within the crate's module tree.
598 The Rust compiler is always invoked with a single source file as input, and always produces a single output crate.
599 The processing of that source file may result in other source files being loaded as modules.
600 Source files have the extension `.rs`.
602 A Rust source file describes a module, the name and
603 location of which -- in the module tree of the current crate -- are defined
604 from outside the source file: either by an explicit `mod_item` in
605 a referencing source file, or by the name of the crate itself.
607 Each source file contains a sequence of zero or more `item` definitions,
608 and may optionally begin with any number of `attributes` that apply to the containing module.
609 Attributes on the anonymous crate module define important metadata that influences
610 the behavior of the compiler.
614 #![crate_id = "projx#2.5"]
616 // Additional metadata attributes
617 #![desc = "Project X"]
619 #![comment = "This is a comment on Project X."]
621 // Specify the output type
622 #![crate_type = "lib"]
625 #![warn(non_camel_case_types)]
628 A crate that contains a `main` function can be compiled to an executable.
629 If a `main` function is present, its return type must be [`unit`](#primitive-types) and it must take no arguments.
631 # Items and attributes
633 Crates contain [items](#items),
634 each of which may have some number of [attributes](#attributes) attached to it.
638 ~~~~ {.notrust .ebnf .gram}
639 item : mod_item | fn_item | type_item | struct_item | enum_item
640 | static_item | trait_item | impl_item | extern_block ;
643 An _item_ is a component of a crate; some module items can be defined in crate
644 files, but most are defined in source files. Items are organized within a
645 crate by a nested set of [modules](#modules). Every crate has a single
646 "outermost" anonymous module; all further items within the crate have
647 [paths](#paths) within the module tree of the crate.
649 Items are entirely determined at compile-time, generally remain fixed during
650 execution, and may reside in read-only memory.
652 There are several kinds of item:
654 * [modules](#modules)
655 * [functions](#functions)
656 * [type definitions](#type-definitions)
657 * [structures](#structures)
658 * [enumerations](#enumerations)
659 * [static items](#static-items)
661 * [implementations](#implementations)
663 Some items form an implicit scope for the declaration of sub-items. In other
664 words, within a function or module, declarations of items can (in many cases)
665 be mixed with the statements, control blocks, and similar artifacts that
666 otherwise compose the item body. The meaning of these scoped items is the same
667 as if the item was declared outside the scope -- it is still a static item --
668 except that the item's *path name* within the module namespace is qualified by
669 the name of the enclosing item, or is private to the enclosing item (in the
671 The grammar specifies the exact locations in which sub-item declarations may appear.
675 All items except modules may be *parameterized* by type. Type parameters are
676 given as a comma-separated list of identifiers enclosed in angle brackets
677 (`<...>`), after the name of the item and before its definition.
678 The type parameters of an item are considered "part of the name", not part of the type of the item.
679 A referencing [path](#paths) must (in principle) provide type arguments as a list of comma-separated types enclosed within angle brackets, in order to refer to the type-parameterized item.
680 In practice, the type-inference system can usually infer such argument types from context.
681 There are no general type-parametric types, only type-parametric items.
682 That is, Rust has no notion of type abstraction: there are no first-class "forall" types.
686 ~~~~ {.notrust .ebnf .gram}
687 mod_item : "mod" ident ( ';' | '{' mod '}' );
688 mod : [ view_item | item ] * ;
691 A module is a container for zero or more [view items](#view-items) and zero or
692 more [items](#items). The view items manage the visibility of the items
693 defined within the module, as well as the visibility of names from outside the
694 module when referenced from inside the module.
696 A _module item_ is a module, surrounded in braces, named, and prefixed with
697 the keyword `mod`. A module item introduces a new, named module into the tree
698 of modules making up a crate. Modules can nest arbitrarily.
700 An example of a module:
704 type Complex = (f64, f64);
705 fn sin(f: f64) -> f64 {
709 fn cos(f: f64) -> f64 {
713 fn tan(f: f64) -> f64 {
720 Modules and types share the same namespace.
721 Declaring a named type that has the same name as a module in scope is forbidden:
722 that is, a type definition, trait, struct, enumeration, or type parameter
723 can't shadow the name of a module in scope, or vice versa.
725 A module without a body is loaded from an external file, by default with the same
726 name as the module, plus the `.rs` extension.
727 When a nested submodule is loaded from an external file,
728 it is loaded from a subdirectory path that mirrors the module hierarchy.
731 // Load the `vec` module from `vec.rs`
735 // Load the `local_data` module from `task/local_data.rs`
740 The directories and files used for loading external file modules can be influenced
741 with the `path` attribute.
744 #[path = "task_files"]
746 // Load the `local_data` module from `task_files/tls.rs`
754 ~~~~ {.notrust .ebnf .gram}
755 view_item : extern_crate_decl | use_decl ;
758 A view item manages the namespace of a module.
759 View items do not define new items, but rather, simply change other items' visibility.
760 There are several kinds of view item:
762 * [`extern crate` declarations](#extern-crate-declarations)
763 * [`use` declarations](#use-declarations)
765 ##### Extern crate declarations
767 ~~~~ {.notrust .ebnf .gram}
768 extern_crate_decl : "extern" "crate" ident [ '(' link_attrs ')' ] ? [ '=' string_lit ] ? ;
769 link_attrs : link_attr [ ',' link_attrs ] + ;
770 link_attr : ident '=' literal ;
773 An _`extern crate` declaration_ specifies a dependency on an external crate.
774 The external crate is then bound into the declaring scope as the `ident` provided
775 in the `extern_crate_decl`.
777 The external crate is resolved to a specific `soname` at compile time, and a
778 runtime linkage requirement to that `soname` is passed to the linker for
779 loading at runtime. The `soname` is resolved at compile time by scanning the
780 compiler's library path and matching the optional `crateid` provided as a string literal
781 against the `crateid` attributes that were declared on the external crate when
782 it was compiled. If no `crateid` is provided, a default `name` attribute is
783 assumed, equal to the `ident` given in the `extern_crate_decl`.
785 Four examples of `extern crate` declarations:
790 extern crate std; // equivalent to: extern crate std = "std";
792 extern crate ruststd = "std"; // linking to 'std' under another name
794 extern crate foo = "some/where/rust-foo#foo:1.0"; // a full crate ID for external tools
797 ##### Use declarations
799 ~~~~ {.notrust .ebnf .gram}
800 use_decl : "pub" ? "use" ident [ '=' path
803 path_glob : ident [ "::" path_glob ] ?
805 | '{' ident [ ',' ident ] * '}' ;
808 A _use declaration_ creates one or more local name bindings synonymous
809 with some other [path](#paths).
810 Usually a `use` declaration is used to shorten the path required to refer to a
811 module item. These declarations may appear at the top of [modules](#modules) and
814 *Note*: Unlike in many languages,
815 `use` declarations in Rust do *not* declare linkage dependency with external crates.
816 Rather, [`extern crate` declarations](#extern-crate-declarations) declare linkage dependencies.
818 Use declarations support a number of convenient shortcuts:
820 * Rebinding the target name as a new local name, using the syntax `use x = p::q::r;`.
821 * Simultaneously binding a list of paths differing only in their final element,
822 using the glob-like brace syntax `use a::b::{c,d,e,f};`
823 * Binding all paths matching a given prefix, using the asterisk wildcard syntax `use a::b::*;`
825 An example of `use` declarations:
828 use std::iter::range_step;
829 use std::option::{Some, None};
834 // Equivalent to 'std::iter::range_step(0, 10, 2);'
835 range_step(0, 10, 2);
837 // Equivalent to 'foo(~[std::option::Some(1.0), std::option::None]);'
838 foo(~[Some(1.0), None]);
842 Like items, `use` declarations are private to the containing module, by default.
843 Also like items, a `use` declaration can be public, if qualified by the `pub` keyword.
844 Such a `use` declaration serves to _re-export_ a name.
845 A public `use` declaration can therefore _redirect_ some public name to a different target definition:
846 even a definition with a private canonical path, inside a different module.
847 If a sequence of such redirections form a cycle or cannot be resolved unambiguously,
848 they represent a compile-time error.
850 An example of re-exporting:
855 pub use quux::foo::*;
864 In this example, the module `quux` re-exports all of the public names defined in `foo`.
866 Also note that the paths contained in `use` items are relative to the crate root.
867 So, in the previous example, the `use` refers to `quux::foo::*`, and not simply to `foo::*`.
868 This also means that top-level module declarations should be at the crate root if direct usage
869 of the declared modules within `use` items is desired. It is also possible to use `self` and `super`
870 at the beginning of a `use` item to refer to the current and direct parent modules respectively.
871 All rules regarding accessing declared modules in `use` declarations applies to both module declarations
872 and `extern crate` declarations.
874 An example of what will and will not work for `use` items:
877 # #![allow(unused_imports)]
878 use foo::native::start; // good: foo is at the root of the crate
879 use foo::baz::foobaz; // good: foo is at the root of the crate
884 use foo::native::start; // good: foo is at crate root
885 // use native::start; // bad: native is not at the crate root
886 use self::baz::foobaz; // good: self refers to module 'foo'
887 use foo::bar::foobar; // good: foo is at crate root
894 use super::bar::foobar; // good: super refers to module 'foo'
904 A _function item_ defines a sequence of [statements](#statements) and an optional final [expression](#expressions), along with a name and a set of parameters.
905 Functions are declared with the keyword `fn`.
906 Functions declare a set of *input* [*slots*](#memory-slots) as parameters, through which the caller passes arguments into the function, and an *output* [*slot*](#memory-slots) through which the function passes results back to the caller.
908 A function may also be copied into a first class *value*, in which case the
909 value has the corresponding [*function type*](#function-types), and can be
910 used otherwise exactly as a function item (with a minor additional cost of
911 calling the function indirectly).
913 Every control path in a function logically ends with a `return` expression or a
914 diverging expression. If the outermost block of a function has a
915 value-producing expression in its final-expression position, that expression
916 is interpreted as an implicit `return` expression applied to the
919 An example of a function:
922 fn add(x: int, y: int) -> int {
927 As with `let` bindings, function arguments are irrefutable patterns,
928 so any pattern that is valid in a let binding is also valid as an argument.
931 fn first((value, _): (int, int)) -> int { value }
935 #### Generic functions
937 A _generic function_ allows one or more _parameterized types_ to
938 appear in its signature. Each type parameter must be explicitly
939 declared, in an angle-bracket-enclosed, comma-separated list following
943 fn iter<T>(seq: &[T], f: |T|) {
944 for elt in seq.iter() { f(elt); }
946 fn map<T, U>(seq: &[T], f: |T| -> U) -> ~[U] {
948 for elt in seq.iter() { acc.push(f(elt)); }
953 Inside the function signature and body, the name of the type parameter
954 can be used as a type name.
956 When a generic function is referenced, its type is instantiated based
957 on the context of the reference. For example, calling the `iter`
958 function defined above on `[1, 2]` will instantiate type parameter `T`
959 with `int`, and require the closure parameter to have type
962 The type parameters can also be explicitly supplied in a trailing
963 [path](#paths) component after the function name. This might be necessary
964 if there is not sufficient context to determine the type parameters. For
965 example, `mem::size_of::<u32>() == 4`.
967 Since a parameter type is opaque to the generic function, the set of
968 operations that can be performed on it is limited. Values of parameter
969 type can only be moved, not copied.
972 fn id<T>(x: T) -> T { x }
975 Similarly, [trait](#traits) bounds can be specified for type
976 parameters to allow methods with that trait to be called on values
982 Unsafe operations are those that potentially violate the memory-safety guarantees of Rust's static semantics.
984 The following language level features cannot be used in the safe subset of Rust:
986 - Dereferencing a [raw pointer](#pointer-types).
987 - Calling an unsafe function (including an intrinsic or foreign function).
989 ##### Unsafe functions
991 Unsafe functions are functions that are not safe in all contexts and/or for all possible inputs.
992 Such a function must be prefixed with the keyword `unsafe`.
996 A block of code can also be prefixed with the `unsafe` keyword, to permit calling `unsafe` functions
997 or dereferencing raw pointers within a safe function.
999 When a programmer has sufficient conviction that a sequence of potentially unsafe operations is
1000 actually safe, they can encapsulate that sequence (taken as a whole) within an `unsafe` block. The
1001 compiler will consider uses of such code safe, in the surrounding context.
1003 Unsafe blocks are used to wrap foreign libraries, make direct use of hardware or implement features
1004 not directly present in the language. For example, Rust provides the language features necessary to
1005 implement memory-safe concurrency in the language but the implementation of tasks and message
1006 passing is in the standard library.
1008 Rust's type system is a conservative approximation of the dynamic safety requirements, so in some
1009 cases there is a performance cost to using safe code. For example, a doubly-linked list is not a
1010 tree structure and can only be represented with managed or reference-counted pointers in safe code.
1011 By using `unsafe` blocks to represent the reverse links as raw pointers, it can be implemented with
1012 only owned pointers.
1014 ##### Behavior considered unsafe
1016 This is a list of behavior which is forbidden in all Rust code. Type checking provides the guarantee
1017 that these issues are never caused by safe code. An `unsafe` block or function is responsible for
1018 never invoking this behaviour or exposing an API making it possible for it to occur in safe code.
1021 * Dereferencing a null/dangling raw pointer
1022 * Mutating an immutable value/reference
1023 * Reads of [undef](http://llvm.org/docs/LangRef.html#undefined-values) (uninitialized) memory
1024 * Breaking the [pointer aliasing rules](http://llvm.org/docs/LangRef.html#pointer-aliasing-rules)
1025 with raw pointers (a subset of the rules used by C)
1026 * Invoking undefined behavior via compiler intrinsics:
1027 * Indexing outside of the bounds of an object with `std::ptr::offset` (`offset` intrinsic), with
1028 the exception of one byte past the end which is permitted.
1029 * Using `std::ptr::copy_nonoverlapping_memory` (`memcpy32`/`memcpy64` instrinsics) on
1031 * Invalid values in primitive types, even in private fields/locals:
1032 * Dangling/null pointers in non-raw pointers, or slices
1033 * A value other than `false` (0) or `true` (1) in a `bool`
1034 * A discriminant in an `enum` not included in the type definition
1035 * A value in a `char` which is a surrogate or above `char::MAX`
1036 * non-UTF-8 byte sequences in a `str`
1038 ##### Behaviour not considered unsafe
1040 This is a list of behaviour not considered *unsafe* in Rust terms, but that may be undesired.
1043 * Reading data from private fields (`std::repr`, `format!("{:?}", x)`)
1044 * Leaks due to reference count cycles, even in the global heap
1045 * Exiting without calling destructors
1047 * Accessing/modifying the file system
1048 * Unsigned integer overflow (well-defined as wrapping)
1049 * Signed integer overflow (well-defined as two's complement representation wrapping)
1051 #### Diverging functions
1053 A special kind of function can be declared with a `!` character where the
1054 output slot type would normally be. For example:
1057 fn my_err(s: &str) -> ! {
1063 We call such functions "diverging" because they never return a value to the
1064 caller. Every control path in a diverging function must end with a
1065 `fail!()` or a call to another diverging function on every
1066 control path. The `!` annotation does *not* denote a type. Rather, the result
1067 type of a diverging function is a special type called $\bot$ ("bottom") that
1068 unifies with any type. Rust has no syntax for $\bot$.
1070 It might be necessary to declare a diverging function because as mentioned
1071 previously, the typechecker checks that every control path in a function ends
1072 with a [`return`](#return-expressions) or diverging expression. So, if `my_err`
1073 were declared without the `!` annotation, the following code would not
1077 # fn my_err(s: &str) -> ! { fail!() }
1079 fn f(i: int) -> int {
1084 my_err("Bad number!");
1089 This will not compile without the `!` annotation on `my_err`,
1090 since the `else` branch of the conditional in `f` does not return an `int`,
1091 as required by the signature of `f`.
1092 Adding the `!` annotation to `my_err` informs the typechecker that,
1093 should control ever enter `my_err`, no further type judgments about `f` need to hold,
1094 since control will never resume in any context that relies on those judgments.
1095 Thus the return type on `f` only needs to reflect the `if` branch of the conditional.
1098 #### Extern functions
1100 Extern functions are part of Rust's foreign function interface,
1101 providing the opposite functionality to [external blocks](#external-blocks).
1102 Whereas external blocks allow Rust code to call foreign code,
1103 extern functions with bodies defined in Rust code _can be called by foreign
1104 code_. They are defined in the same way as any other Rust function,
1105 except that they have the `extern` modifier.
1108 // Declares an extern fn, the ABI defaults to "C"
1109 extern fn new_vec() -> ~[int] { ~[] }
1111 // Declares an extern fn with "stdcall" ABI
1112 extern "stdcall" fn new_vec_stdcall() -> ~[int] { ~[] }
1115 Unlike normal functions, extern fns have an `extern "ABI" fn()`.
1116 This is the same type as the functions declared in an extern
1120 # extern fn new_vec() -> ~[int] { ~[] }
1121 let fptr: extern "C" fn() -> ~[int] = new_vec;
1124 Extern functions may be called directly from Rust code as Rust uses large,
1125 contiguous stack segments like C.
1127 ### Type definitions
1129 A _type definition_ defines a new name for an existing [type](#types). Type
1130 definitions are declared with the keyword `type`. Every value has a single,
1131 specific type; the type-specified aspects of a value include:
1133 * Whether the value is composed of sub-values or is indivisible.
1134 * Whether the value represents textual or numerical information.
1135 * Whether the value represents integral or floating-point information.
1136 * The sequence of memory operations required to access the value.
1137 * The [kind](#type-kinds) of the type.
1139 For example, the type `(u8, u8)` defines the set of immutable values that are composite pairs,
1140 each containing two unsigned 8-bit integers accessed by pattern-matching and laid out in memory with the `x` component preceding the `y` component.
1144 A _structure_ is a nominal [structure type](#structure-types) defined with the keyword `struct`.
1146 An example of a `struct` item and its use:
1149 struct Point {x: int, y: int}
1150 let p = Point {x: 10, y: 11};
1154 A _tuple structure_ is a nominal [tuple type](#tuple-types), also defined with the keyword `struct`.
1158 struct Point(int, int);
1159 let p = Point(10, 11);
1160 let px: int = match p { Point(x, _) => x };
1163 A _unit-like struct_ is a structure without any fields, defined by leaving off the list of fields entirely.
1164 Such types will have a single value, just like the [unit value `()`](#unit-and-boolean-literals) of the unit type.
1169 let c = [Cookie, Cookie, Cookie, Cookie];
1174 An _enumeration_ is a simultaneous definition of a nominal [enumerated type](#enumerated-types) as well as a set of *constructors*,
1175 that can be used to create or pattern-match values of the corresponding enumerated type.
1177 Enumerations are declared with the keyword `enum`.
1179 An example of an `enum` item and its use:
1187 let mut a: Animal = Dog;
1191 Enumeration constructors can have either named or unnamed fields:
1196 Cat { name: ~str, weight: f64 }
1199 let mut a: Animal = Dog(~"Cocoa", 37.2);
1200 a = Cat{ name: ~"Spotty", weight: 2.7 };
1203 In this example, `Cat` is a _struct-like enum variant_,
1204 whereas `Dog` is simply called an enum variant.
1208 ~~~~ {.notrust .ebnf .gram}
1209 static_item : "static" ident ':' type '=' expr ';' ;
1212 A *static item* is a named _constant value_ stored in the global data section of a crate.
1213 Immutable static items are stored in the read-only data section.
1214 The constant value bound to a static item is, like all constant values, evaluated at compile time.
1215 Static items have the `static` lifetime, which outlives all other lifetimes in a Rust program.
1216 Static items are declared with the `static` keyword.
1217 A static item must have a _constant expression_ giving its definition.
1219 Static items must be explicitly typed.
1220 The type may be ```bool```, ```char```, a number, or a type derived from those primitive types.
1221 The derived types are references with the `static` lifetime,
1222 fixed-size arrays, tuples, and structs.
1225 static BIT1: uint = 1 << 0;
1226 static BIT2: uint = 1 << 1;
1228 static BITS: [uint, ..2] = [BIT1, BIT2];
1229 static STRING: &'static str = "bitstring";
1231 struct BitsNStrings<'a> {
1232 mybits: [uint, ..2],
1236 static bits_n_strings: BitsNStrings<'static> = BitsNStrings {
1242 #### Mutable statics
1244 If a static item is declared with the ```mut``` keyword, then it is allowed to
1245 be modified by the program. One of Rust's goals is to make concurrency bugs hard
1246 to run into, and this is obviously a very large source of race conditions or
1247 other bugs. For this reason, an ```unsafe``` block is required when either
1248 reading or writing a mutable static variable. Care should be taken to ensure
1249 that modifications to a mutable static are safe with respect to other tasks
1250 running in the same process.
1252 Mutable statics are still very useful, however. They can be used with C
1253 libraries and can also be bound from C libraries (in an ```extern``` block).
1256 # fn atomic_add(_: &mut uint, _: uint) -> uint { 2 }
1258 static mut LEVELS: uint = 0;
1260 // This violates the idea of no shared state, and this doesn't internally
1261 // protect against races, so this function is `unsafe`
1262 unsafe fn bump_levels_unsafe1() -> uint {
1268 // Assuming that we have an atomic_add function which returns the old value,
1269 // this function is "safe" but the meaning of the return value may not be what
1270 // callers expect, so it's still marked as `unsafe`
1271 unsafe fn bump_levels_unsafe2() -> uint {
1272 return atomic_add(&mut LEVELS, 1);
1278 A _trait_ describes a set of method types.
1280 Traits can include default implementations of methods,
1281 written in terms of some unknown [`self` type](#self-types);
1282 the `self` type may either be completely unspecified,
1283 or constrained by some other trait.
1285 Traits are implemented for specific types through separate [implementations](#implementations).
1288 # type Surface = int;
1289 # type BoundingBox = int;
1292 fn draw(&self, Surface);
1293 fn bounding_box(&self) -> BoundingBox;
1297 This defines a trait with two methods.
1298 All values that have [implementations](#implementations) of this trait in scope can have their `draw` and `bounding_box` methods called,
1299 using `value.bounding_box()` [syntax](#method-call-expressions).
1301 Type parameters can be specified for a trait to make it generic.
1302 These appear after the trait name, using the same syntax used in [generic functions](#generic-functions).
1306 fn len(&self) -> uint;
1307 fn elt_at(&self, n: uint) -> T;
1308 fn iter(&self, |T|);
1312 Generic functions may use traits as _bounds_ on their type parameters.
1313 This will have two effects: only types that have the trait may instantiate the parameter,
1314 and within the generic function,
1315 the methods of the trait can be called on values that have the parameter's type.
1319 # type Surface = int;
1320 # trait Shape { fn draw(&self, Surface); }
1322 fn draw_twice<T: Shape>(surface: Surface, sh: T) {
1328 Traits also define an [object type](#object-types) with the same name as the trait.
1329 Values of this type are created by [casting](#type-cast-expressions) pointer values
1330 (pointing to a type for which an implementation of the given trait is in scope)
1331 to pointers to the trait name, used as a type.
1335 # impl Shape for int { }
1338 let myshape: ~Shape = ~mycircle as ~Shape;
1341 The resulting value is a managed box containing the value that was cast,
1342 along with information that identifies the methods of the implementation that was used.
1343 Values with a trait type can have [methods called](#method-call-expressions) on them,
1344 for any method in the trait,
1345 and can be used to instantiate type parameters that are bounded by the trait.
1347 Trait methods may be static,
1348 which means that they lack a `self` argument.
1349 This means that they can only be called with function call syntax (`f(x)`)
1350 and not method call syntax (`obj.f()`).
1351 The way to refer to the name of a static method is to qualify it with the trait name,
1352 treating the trait name like a module.
1357 fn from_int(n: int) -> Self;
1360 fn from_int(n: int) -> f64 { n as f64 }
1362 let x: f64 = Num::from_int(42);
1365 Traits may inherit from other traits. For example, in
1368 trait Shape { fn area() -> f64; }
1369 trait Circle : Shape { fn radius() -> f64; }
1372 the syntax `Circle : Shape` means that types that implement `Circle` must also have an implementation for `Shape`.
1373 Multiple supertraits are separated by spaces, `trait Circle : Shape Eq { }`.
1374 In an implementation of `Circle` for a given type `T`, methods can refer to `Shape` methods,
1375 since the typechecker checks that any type with an implementation of `Circle` also has an implementation of `Shape`.
1377 In type-parameterized functions,
1378 methods of the supertrait may be called on values of subtrait-bound type parameters.
1379 Referring to the previous example of `trait Circle : Shape`:
1382 # trait Shape { fn area(&self) -> f64; }
1383 # trait Circle : Shape { fn radius(&self) -> f64; }
1384 fn radius_times_area<T: Circle>(c: T) -> f64 {
1385 // `c` is both a Circle and a Shape
1386 c.radius() * c.area()
1390 Likewise, supertrait methods may also be called on trait objects.
1393 # trait Shape { fn area(&self) -> f64; }
1394 # trait Circle : Shape { fn radius(&self) -> f64; }
1395 # impl Shape for int { fn area(&self) -> f64 { 0.0 } }
1396 # impl Circle for int { fn radius(&self) -> f64 { 0.0 } }
1399 let mycircle: Circle = ~mycircle as ~Circle;
1400 let nonsense = mycircle.radius() * mycircle.area();
1405 An _implementation_ is an item that implements a [trait](#traits) for a specific type.
1407 Implementations are defined with the keyword `impl`.
1410 # struct Point {x: f64, y: f64};
1411 # type Surface = int;
1412 # struct BoundingBox {x: f64, y: f64, width: f64, height: f64};
1413 # trait Shape { fn draw(&self, Surface); fn bounding_box(&self) -> BoundingBox; }
1414 # fn do_draw_circle(s: Surface, c: Circle) { }
1421 impl Shape for Circle {
1422 fn draw(&self, s: Surface) { do_draw_circle(s, *self); }
1423 fn bounding_box(&self) -> BoundingBox {
1424 let r = self.radius;
1425 BoundingBox{x: self.center.x - r, y: self.center.y - r,
1426 width: 2.0 * r, height: 2.0 * r}
1431 It is possible to define an implementation without referring to a trait.
1432 The methods in such an implementation can only be used
1433 as direct calls on the values of the type that the implementation targets.
1434 In such an implementation, the trait type and `for` after `impl` are omitted.
1435 Such implementations are limited to nominal types (enums, structs),
1436 and the implementation must appear in the same module or a sub-module as the `self` type.
1438 When a trait _is_ specified in an `impl`,
1439 all methods declared as part of the trait must be implemented,
1440 with matching types and type parameter counts.
1442 An implementation can take type parameters,
1443 which can be different from the type parameters taken by the trait it implements.
1444 Implementation parameters are written after the `impl` keyword.
1449 impl<T> Seq<T> for ~[T] {
1452 impl Seq<bool> for u32 {
1453 /* Treat the integer as a sequence of bits */
1459 ~~~~ {.notrust .ebnf .gram}
1460 extern_block_item : "extern" '{' extern_block '}' ;
1461 extern_block : [ foreign_fn ] * ;
1464 External blocks form the basis for Rust's foreign function interface.
1465 Declarations in an external block describe symbols
1466 in external, non-Rust libraries.
1468 Functions within external blocks
1469 are declared in the same way as other Rust functions,
1470 with the exception that they may not have a body
1471 and are instead terminated by a semicolon.
1475 use libc::{c_char, FILE};
1478 fn fopen(filename: *c_char, mode: *c_char) -> *FILE;
1483 Functions within external blocks may be called by Rust code,
1484 just like functions defined in Rust.
1485 The Rust compiler automatically translates
1486 between the Rust ABI and the foreign ABI.
1488 A number of [attributes](#attributes) control the behavior of external
1491 By default external blocks assume that the library they are calling
1492 uses the standard C "cdecl" ABI. Other ABIs may be specified using
1493 an `abi` string, as shown here:
1496 // Interface to the Windows API
1497 extern "stdcall" { }
1500 The `link` attribute allows the name of the library to be specified. When
1501 specified the compiler will attempt to link against the native library of the
1505 #[link(name = "crypto")]
1509 The type of a function declared in an extern block is `extern "abi" fn(A1,
1510 ..., An) -> R`, where `A1...An` are the declared types of its arguments and
1511 `R` is the declared return type.
1513 ## Visibility and Privacy
1515 These two terms are often used interchangeably, and what they are attempting to
1516 convey is the answer to the question "Can this item be used at this location?"
1518 Rust's name resolution operates on a global hierarchy of namespaces. Each level
1519 in the hierarchy can be thought of as some item. The items are one of those
1520 mentioned above, but also include external crates. Declaring or defining a new
1521 module can be thought of as inserting a new tree into the hierarchy at the
1522 location of the definition.
1524 To control whether interfaces can be used across modules, Rust checks each use
1525 of an item to see whether it should be allowed or not. This is where privacy
1526 warnings are generated, or otherwise "you used a private item of another module
1527 and weren't allowed to."
1529 By default, everything in rust is *private*, with one exception. Enum variants
1530 in a `pub` enum are also public by default. You are allowed to alter this
1531 default visibility with the `priv` keyword. When an item is declared as `pub`,
1532 it can be thought of as being accessible to the outside world. For example:
1536 // Declare a private struct
1539 // Declare a public struct with a private field
1544 // Declare a public enum with public and private variants
1546 PubliclyAccessibleState,
1547 priv PrivatelyAccessibleState
1551 With the notion of an item being either public or private, Rust allows item
1552 accesses in two cases:
1554 1. If an item is public, then it can be used externally through any of its
1556 2. If an item is private, it may be accessed by the current module and its
1559 These two cases are surprisingly powerful for creating module hierarchies
1560 exposing public APIs while hiding internal implementation details. To help
1561 explain, here's a few use cases and what they would entail.
1563 * A library developer needs to expose functionality to crates which link against
1564 their library. As a consequence of the first case, this means that anything
1565 which is usable externally must be `pub` from the root down to the destination
1566 item. Any private item in the chain will disallow external accesses.
1568 * A crate needs a global available "helper module" to itself, but it doesn't
1569 want to expose the helper module as a public API. To accomplish this, the root
1570 of the crate's hierarchy would have a private module which then internally has
1571 a "public api". Because the entire crate is a descendant of the root, then the
1572 entire local crate can access this private module through the second case.
1574 * When writing unit tests for a module, it's often a common idiom to have an
1575 immediate child of the module to-be-tested named `mod test`. This module could
1576 access any items of the parent module through the second case, meaning that
1577 internal implementation details could also be seamlessly tested from the child
1580 In the second case, it mentions that a private item "can be accessed" by the
1581 current module and its descendants, but the exact meaning of accessing an item
1582 depends on what the item is. Accessing a module, for example, would mean looking
1583 inside of it (to import more items). On the other hand, accessing a function
1584 would mean that it is invoked. Additionally, path expressions and import
1585 statements are considered to access an item in the sense that the
1586 import/expression is only valid if the destination is in the current visibility
1589 Here's an example of a program which exemplifies the three cases outlined above.
1592 // This module is private, meaning that no external crate can access this
1593 // module. Because it is private at the root of this current crate, however, any
1594 // module in the crate may access any publicly visible item in this module.
1595 mod crate_helper_module {
1597 // This function can be used by anything in the current crate
1598 pub fn crate_helper() {}
1600 // This function *cannot* be used by anything else in the crate. It is not
1601 // publicly visible outside of the `crate_helper_module`, so only this
1602 // current module and its descendants may access it.
1603 fn implementation_detail() {}
1606 // This function is "public to the root" meaning that it's available to external
1607 // crates linking against this one.
1608 pub fn public_api() {}
1610 // Similarly to 'public_api', this module is public so external crates may look
1613 use crate_helper_module;
1615 pub fn my_method() {
1616 // Any item in the local crate may invoke the helper module's public
1617 // interface through a combination of the two rules above.
1618 crate_helper_module::crate_helper();
1621 // This function is hidden to any module which is not a descendant of
1623 fn my_implementation() {}
1629 fn test_my_implementation() {
1630 // Because this module is a descendant of `submodule`, it's allowed
1631 // to access private items inside of `submodule` without a privacy
1633 super::my_implementation();
1641 For a rust program to pass the privacy checking pass, all paths must be valid
1642 accesses given the two rules above. This includes all use statements,
1643 expressions, types, etc.
1645 ### Re-exporting and Visibility
1647 Rust allows publicly re-exporting items through a `pub use` directive. Because
1648 this is a public directive, this allows the item to be used in the current
1649 module through the rules above. It essentially allows public access into the
1650 re-exported item. For example, this program is valid:
1653 pub use api = self::implementation;
1655 mod implementation {
1662 This means that any external crate referencing `implementation::f` would receive
1663 a privacy violation, while the path `api::f` would be allowed.
1665 When re-exporting a private item, it can be thought of as allowing the "privacy
1666 chain" being short-circuited through the reexport instead of passing through the
1667 namespace hierarchy as it normally would.
1669 ### Glob imports and Visibility
1671 Currently glob imports are considered an "experimental" language feature. For
1672 sanity purpose along with helping the implementation, glob imports will only
1673 import public items from their destination, not private items.
1675 > **Note:** This is subject to change, glob exports may be removed entirely or
1676 > they could possibly import private items for a privacy error to later be
1677 > issued if the item is used.
1681 ~~~~ {.notrust .ebnf .gram}
1682 attribute : '#' '!' ? '[' attr_list ']' ;
1683 attr_list : attr [ ',' attr_list ]* ;
1684 attr : ident [ '=' literal
1685 | '(' attr_list ')' ] ? ;
1688 Static entities in Rust -- crates, modules and items -- may have _attributes_
1689 applied to them. Attributes in Rust are modeled on Attributes in ECMA-335,
1690 with the syntax coming from ECMA-334 (C#). An attribute is a general,
1691 free-form metadatum that is interpreted according to name, convention, and
1692 language and compiler version. Attributes may appear as any of:
1694 * A single identifier, the attribute name
1695 * An identifier followed by the equals sign '=' and a literal, providing a
1697 * An identifier followed by a parenthesized list of sub-attribute arguments
1699 Attributes with a bang ("!") after the hash ("#") apply to the item that the
1700 attribute is declared within. Attributes that do not have a bang after the
1701 hash apply to the item that follows the attribute.
1703 An example of attributes:
1706 // General metadata applied to the enclosing module or crate.
1709 // A function marked as a unit test
1715 // A conditionally-compiled module
1716 #[cfg(target_os="linux")]
1721 // A lint attribute used to suppress a warning/error
1722 #[allow(non_camel_case_types)]
1726 > **Note:** At some point in the future, the compiler will distinguish between
1727 > language-reserved and user-available attributes. Until then, there is
1728 > effectively no difference between an attribute handled by a loadable syntax
1729 > extension and the compiler.
1731 ### Crate-only attributes
1733 - `crate_id` - specify the this crate's crate ID.
1734 - `crate_type` - see [linkage](#linkage).
1735 - `feature` - see [compiler features](#compiler-features).
1736 - `no_main` - disable emitting the `main` symbol. Useful when some other
1737 object being linked to defines `main`.
1738 - `no_start` - disable linking to the `native` crate, which specifies the
1739 "start" language item.
1740 - `no_std` - disable linking to the `std` crate.
1742 ### Module-only attributes
1744 - `macro_escape` - macros defined in this module will be visible in the
1745 module's parent, after this module has been included.
1746 - `no_implicit_prelude` - disable injecting `use std::prelude::*` in this
1748 - `path` - specifies the file to load the module from. `#[path="foo.rs"] mod
1749 bar;` is equivalent to `mod bar { /* contents of foo.rs */ }`. The path is
1750 taken relative to the directory that the current module is in.
1752 ### Function-only attributes
1754 - `macro_registrar` - when using loadable syntax extensions, mark this
1755 function as the registration point for the current crate's syntax
1757 - `main` - indicates that this function should be passed to the entry point,
1758 rather than the function in the crate root named `main`.
1759 - `start` - indicates that this function should be used as the entry point,
1760 overriding the "start" language item. See the "start" [language
1761 item](#language-items) for more details.
1763 ### Static-only attributes
1765 - `address_insignificant` - references to this static may alias with
1766 references to other statics, potentially of unrelated type.
1767 - `thread_local` - on a `static mut`, this signals that the value of this
1768 static may change depending on the current thread. The exact consequences of
1769 this are implementation-defined.
1773 On an `extern` block, the following attributes are interpreted:
1775 - `link_args` - specify arguments to the linker, rather than just the library
1776 name and type. This is feature gated and the exact behavior is
1777 implementation-defined (due to variety of linker invocation syntax).
1778 - `link` - indicate that a native library should be linked to for the
1779 declarations in this block to be linked correctly. See [external
1780 blocks](#external-blocks)
1782 On declarations inside an `extern` block, the following attributes are
1785 - `link_name` - the name of the symbol that this function or static should be
1787 - `linkage` - on a static, this specifies the [linkage
1788 type](http://llvm.org/docs/LangRef.html#linkage-types).
1790 ### Miscellaneous attributes
1792 - `link_section` - on statics and functions, this specifies the section of the
1793 object file that this item's contents will be placed into.
1794 - `macro_export` - export a macro for cross-crate usage.
1795 - `no_mangle` - on any item, do not apply the standard name mangling. Set the
1796 symbol for this item to its identifier.
1797 - `packed` - on structs or enums, eliminate any padding that would be used to
1799 - `repr` - on C-like enums, this sets the underlying type used for
1800 representation. Useful for FFI. Takes one argument, which is the primitive
1801 type this enum should be represented for, or `C`, which specifies that it
1802 should be the default `enum` size of the C ABI for that platform. Note that
1803 enum representation in C is undefined, and this may be incorrect when the C
1804 code is compiled with certain flags.
1805 - `simd` - on certain tuple structs, derive the arithmetic operators, which
1806 lower to the target's SIMD instructions, if any.
1807 - `static_assert` - on statics whose type is `bool`, terminates compilation
1808 with an error if it is not initialized to `true`.
1809 - `unsafe_destructor` - allow implementations of the "drop" language item
1810 where the type it is implemented for does not implement the "send" language
1812 - `unsafe_no_drop_flag` - on structs, remove the flag that prevents
1813 destructors from being run twice. Destructors might be run multiple times on
1814 the same object with this attribute.
1816 ### Conditional compilation
1818 Sometimes one wants to have different compiler outputs from the same code,
1819 depending on build target, such as targeted operating system, or to enable
1822 There are two kinds of configuration options, one that is either defined or not
1823 (`#[cfg(foo)]`), and the other that contains a string that can be checked
1824 against (`#[cfg(bar = "baz")]` (currently only compiler-defined configuration
1825 options can have the latter form).
1828 // The function is only included in the build when compiling for OSX
1829 #[cfg(target_os = "macos")]
1834 // This function is only included when either foo or bar is defined
1837 fn needs_foo_or_bar() {
1841 // This function is only included when compiling for a unixish OS with a 32-bit
1843 #[cfg(unix, target_word_size = "32")]
1844 fn on_32bit_unix() {
1849 This illustrates some conditional compilation can be achieved using the
1850 `#[cfg(...)]` attribute. Note that `#[cfg(foo, bar)]` is a condition that needs
1851 both `foo` and `bar` to be defined while `#[cfg(foo)] #[cfg(bar)]` only needs
1852 one of `foo` and `bar` to be defined (this resembles in the disjunctive normal
1853 form). Additionally, one can reverse a condition by enclosing it in a
1854 `not(...)`, like e. g. `#[cfg(not(target_os = "win32"))]`.
1856 The following configurations must be defined by the implementation:
1858 * `target_arch = "..."`. Target CPU architecture, such as `"x86"`, `"x86_64"`
1859 `"mips"`, or `"arm"`.
1860 * `target_endian = "..."`. Endianness of the target CPU, either `"little"` or
1862 * `target_family = "..."`. Operating system family of the target, e. g.
1863 `"unix"` or `"windows"`. The value of this configuration option is defined as
1864 a configuration itself, like `unix` or `windows`.
1865 * `target_os = "..."`. Operating system of the target, examples include
1866 `"win32"`, `"macos"`, `"linux"`, `"android"` or `"freebsd"`.
1867 * `target_word_size = "..."`. Target word size in bits. This is set to `"32"`
1868 for targets with 32-bit pointers, and likewise set to `"64"` for 64-bit
1870 * `unix`. See `target_family`.
1871 * `windows`. See `target_family`.
1873 ### Lint check attributes
1875 A lint check names a potentially undesirable coding pattern, such as
1876 unreachable code or omitted documentation, for the static entity to
1877 which the attribute applies.
1879 For any lint check `C`:
1881 * `warn(C)` warns about violations of `C` but continues compilation,
1882 * `deny(C)` signals an error after encountering a violation of `C`,
1883 * `allow(C)` overrides the check for `C` so that violations will go
1885 * `forbid(C)` is the same as `deny(C)`, but also forbids changing the lint
1888 The lint checks supported by the compiler can be found via `rustc -W help`,
1889 along with their default settings.
1893 // Missing documentation is ignored here
1894 #[allow(missing_doc)]
1895 pub fn undocumented_one() -> int { 1 }
1897 // Missing documentation signals a warning here
1898 #[warn(missing_doc)]
1899 pub fn undocumented_too() -> int { 2 }
1901 // Missing documentation signals an error here
1902 #[deny(missing_doc)]
1903 pub fn undocumented_end() -> int { 3 }
1907 This example shows how one can use `allow` and `warn` to toggle
1908 a particular check on and off.
1911 #[warn(missing_doc)]
1913 #[allow(missing_doc)]
1915 // Missing documentation is ignored here
1916 pub fn undocumented_one() -> int { 1 }
1918 // Missing documentation signals a warning here,
1919 // despite the allow above.
1920 #[warn(missing_doc)]
1921 pub fn undocumented_two() -> int { 2 }
1924 // Missing documentation signals a warning here
1925 pub fn undocumented_too() -> int { 3 }
1929 This example shows how one can use `forbid` to disallow uses
1930 of `allow` for that lint check.
1933 #[forbid(missing_doc)]
1935 // Attempting to toggle warning signals an error here
1936 #[allow(missing_doc)]
1938 pub fn undocumented_too() -> int { 2 }
1944 Some primitive Rust operations are defined in Rust code, rather than being
1945 implemented directly in C or assembly language. The definitions of these
1946 operations have to be easy for the compiler to find. The `lang` attribute
1947 makes it possible to declare these operations. For example, the `str` module
1948 in the Rust standard library defines the string equality function:
1952 pub fn eq_slice(a: &str, b: &str) -> bool {
1957 The name `str_eq` has a special meaning to the Rust compiler,
1958 and the presence of this definition means that it will use this definition
1959 when generating calls to the string equality function.
1961 A complete list of the built-in language items follows:
1963 #### Built-in Traits
1966 : Able to be sent across task boundaries.
1968 : Has a size known at compile time.
1970 : Types that do not move ownership when used by-value.
1972 : Able to be safely shared between tasks when aliased.
1978 These language items are traits:
1981 : Elements can be added (for example, integers and floats).
1983 : Elements can be subtracted.
1985 : Elements can be multiplied.
1987 : Elements have a division operation.
1989 : Elements have a remainder operation.
1991 : Elements can be negated arithmetically.
1993 : Elements can be negated logically.
1995 : Elements have an exclusive-or operation.
1997 : Elements have a bitwise `and` operation.
1999 : Elements have a bitwise `or` operation.
2001 : Elements have a left shift operation.
2003 : Elements have a right shift operation.
2005 : Elements can be indexed.
2007 : Elements can be compared for equality.
2009 : Elements have a partial ordering.
2011 : `*` can be applied, yielding a reference to another type
2013 : `*` can be applied, yielding a mutable reference to another type
2016 These are functions:
2019 : Compare two strings (`&str`) for equality.
2021 : Compare two owned strings (`~str`) for equality.
2023 : Return a new unique string
2024 containing a copy of the contents of a unique string.
2029 : A type whose contents can be mutated through an immutable reference
2031 : The type returned by the `type_id` intrinsic.
2035 These types help drive the compiler's analysis
2038 : The type parameter should be considered covariant
2039 `contravariant_type`
2040 : The type parameter should be considered contravariant
2042 : The type parameter should be considered invariant
2043 `covariant_lifetime`
2044 : The lifetime parameter should be considered covariant
2045 `contravariant_lifetime`
2046 : The lifetime parameter should be considered contravariant
2047 `invariant_lifetime`
2048 : The lifetime parameter should be considered invariant
2050 : This type does not implement "send", even if eligible
2052 : This type does not implement "copy", even if eligible
2054 : This type does not implement "share", even if eligible
2056 : This type implements "managed"
2059 : Abort the program with an error.
2061 : Abort the program with a bounds check error.
2063 : Allocate memory on the exchange heap.
2065 : Free memory that was allocated on the exchange heap.
2067 : Allocate memory on the managed heap.
2069 : Free memory that was allocated on the managed heap.
2071 > **Note:** This list is likely to become out of date. We should auto-generate it
2072 > from `librustc/middle/lang_items.rs`.
2074 ### Inline attributes
2076 The inline attribute is used to suggest to the compiler to perform an inline
2077 expansion and place a copy of the function in the caller rather than generating
2078 code to call the function where it is defined.
2080 The compiler automatically inlines functions based on internal heuristics.
2081 Incorrectly inlining functions can actually making the program slower, so it
2082 should be used with care.
2084 `#[inline]` and `#[inline(always)]` always causes the function to be serialized
2085 into crate metadata to allow cross-crate inlining.
2087 There are three different types of inline attributes:
2089 * `#[inline]` hints the compiler to perform an inline expansion.
2090 * `#[inline(always)]` asks the compiler to always perform an inline expansion.
2091 * `#[inline(never)]` asks the compiler to never perform an inline expansion.
2095 The `deriving` attribute allows certain traits to be automatically
2096 implemented for data structures. For example, the following will
2097 create an `impl` for the `Eq` and `Clone` traits for `Foo`, the type
2098 parameter `T` will be given the `Eq` or `Clone` constraints for the
2102 #[deriving(Eq, Clone)]
2109 The generated `impl` for `Eq` is equivalent to
2112 # struct Foo<T> { a: int, b: T }
2113 impl<T: Eq> Eq for Foo<T> {
2114 fn eq(&self, other: &Foo<T>) -> bool {
2115 self.a == other.a && self.b == other.b
2118 fn ne(&self, other: &Foo<T>) -> bool {
2119 self.a != other.a || self.b != other.b
2124 Supported traits for `deriving` are:
2126 * Comparison traits: `Eq`, `TotalEq`, `Ord`, `TotalOrd`.
2127 * Serialization: `Encodable`, `Decodable`. These require `serialize`.
2128 * `Clone`, to create `T` from `&T` via a copy.
2129 * `Hash`, to iterate over the bytes in a data type.
2130 * `Rand`, to create a random instance of a data type.
2131 * `Default`, to create an empty instance of a data type.
2132 * `Zero`, to create an zero instance of a numeric data type.
2133 * `FromPrimitive`, to create an instance from a numeric primitive.
2134 * `Show`, to format a value using the `{}` formatter.
2138 One can indicate the stability of an API using the following attributes:
2140 * `deprecated`: This item should no longer be used, e.g. it has been
2141 replaced. No guarantee of backwards-compatibility.
2142 * `experimental`: This item was only recently introduced or is
2143 otherwise in a state of flux. It may change significantly, or even
2144 be removed. No guarantee of backwards-compatibility.
2145 * `unstable`: This item is still under development, but requires more
2146 testing to be considered stable. No guarantee of backwards-compatibility.
2147 * `stable`: This item is considered stable, and will not change
2148 significantly. Guarantee of backwards-compatibility.
2149 * `frozen`: This item is very stable, and is unlikely to
2150 change. Guarantee of backwards-compatibility.
2151 * `locked`: This item will never change unless a serious bug is
2152 found. Guarantee of backwards-compatibility.
2154 These levels are directly inspired by
2155 [Node.js' "stability index"](http://nodejs.org/api/documentation.html).
2157 There are lints for disallowing items marked with certain levels:
2158 `deprecated`, `experimental` and `unstable`; the first two will warn
2159 by default. Items with not marked with a stability are considered to
2160 be unstable for the purposes of the lint. One can give an optional
2161 string that will be displayed when the lint flags the use of an item.
2166 #[deprecated="replaced by `best`"]
2168 // delete everything
2172 // delete fewer things
2181 bad(); // "warning: use of deprecated item: replaced by `best`"
2183 better(); // "warning: use of unmarked item"
2185 best(); // no warning
2189 > **Note:** Currently these are only checked when applied to
2190 > individual functions, structs, methods and enum variants, *not* to
2191 > entire modules, traits, impls or enums themselves.
2193 ### Compiler Features
2195 Certain aspects of Rust may be implemented in the compiler, but they're not
2196 necessarily ready for every-day use. These features are often of "prototype
2197 quality" or "almost production ready", but may not be stable enough to be
2198 considered a full-fleged language feature.
2200 For this reason, Rust recognizes a special crate-level attribute of the form:
2203 #![feature(feature1, feature2, feature3)]
2206 This directive informs the compiler that the feature list: `feature1`,
2207 `feature2`, and `feature3` should all be enabled. This is only recognized at a
2208 crate-level, not at a module-level. Without this directive, all features are
2209 considered off, and using the features will result in a compiler error.
2211 The currently implemented features of the reference compiler are:
2213 * `macro_rules` - The definition of new macros. This does not encompass
2214 macro-invocation, that is always enabled by default, this only
2215 covers the definition of new macros. There are currently
2216 various problems with invoking macros, how they interact with
2217 their environment, and possibly how they are used outside of
2218 location in which they are defined. Macro definitions are
2219 likely to change slightly in the future, so they are currently
2220 hidden behind this feature.
2222 * `globs` - Importing everything in a module through `*`. This is currently a
2223 large source of bugs in name resolution for Rust, and it's not clear
2224 whether this will continue as a feature or not. For these reasons,
2225 the glob import statement has been hidden behind this feature flag.
2227 * `struct_variant` - Structural enum variants (those with named fields). It is
2228 currently unknown whether this style of enum variant is as
2229 fully supported as the tuple-forms, and it's not certain
2230 that this style of variant should remain in the language.
2231 For now this style of variant is hidden behind a feature
2234 * `once_fns` - Onceness guarantees a closure is only executed once. Defining a
2235 closure as `once` is unlikely to be supported going forward. So
2236 they are hidden behind this feature until they are to be removed.
2238 * `managed_boxes` - Usage of `@` pointers is gated due to many
2239 planned changes to this feature. In the past, this has meant
2240 "a GC pointer", but the current implementation uses
2241 reference counting and will likely change drastically over
2242 time. Additionally, the `@` syntax will no longer be used to
2245 * `asm` - The `asm!` macro provides a means for inline assembly. This is often
2246 useful, but the exact syntax for this feature along with its semantics
2247 are likely to change, so this macro usage must be opted into.
2249 * `non_ascii_idents` - The compiler supports the use of non-ascii identifiers,
2250 but the implementation is a little rough around the
2251 edges, so this can be seen as an experimental feature for
2252 now until the specification of identifiers is fully
2255 * `thread_local` - The usage of the `#[thread_local]` attribute is experimental
2256 and should be seen as unstable. This attribute is used to
2257 declare a `static` as being unique per-thread leveraging
2258 LLVM's implementation which works in concert with the kernel
2259 loader and dynamic linker. This is not necessarily available
2260 on all platforms, and usage of it is discouraged (rust
2261 focuses more on task-local data instead of thread-local
2264 * `link_args` - This attribute is used to specify custom flags to the linker,
2265 but usage is strongly discouraged. The compiler's usage of the
2266 system linker is not guaranteed to continue in the future, and
2267 if the system linker is not used then specifying custom flags
2268 doesn't have much meaning.
2270 If a feature is promoted to a language feature, then all existing programs will
2271 start to receive compilation warnings about #[feature] directives which enabled
2272 the new feature (because the directive is no longer necessary). However, if
2273 a feature is decided to be removed from the language, errors will be issued (if
2274 there isn't a parser error first). The directive in this case is no longer
2275 necessary, and it's likely that existing code will break if the feature isn't
2278 If a unknown feature is found in a directive, it results in a compiler error. An
2279 unknown feature is one which has never been recognized by the compiler.
2281 # Statements and expressions
2283 Rust is _primarily_ an expression language. This means that most forms of
2284 value-producing or effect-causing evaluation are directed by the uniform
2285 syntax category of _expressions_. Each kind of expression can typically _nest_
2286 within each other kind of expression, and rules for evaluation of expressions
2287 involve specifying both the value produced by the expression and the order in
2288 which its sub-expressions are themselves evaluated.
2290 In contrast, statements in Rust serve _mostly_ to contain and explicitly
2291 sequence expression evaluation.
2295 A _statement_ is a component of a block, which is in turn a component of an
2296 outer [expression](#expressions) or [function](#functions).
2298 Rust has two kinds of statement:
2299 [declaration statements](#declaration-statements) and
2300 [expression statements](#expression-statements).
2302 ### Declaration statements
2304 A _declaration statement_ is one that introduces one or more *names* into the enclosing statement block.
2305 The declared names may denote new slots or new items.
2307 #### Item declarations
2309 An _item declaration statement_ has a syntactic form identical to an
2310 [item](#items) declaration within a module. Declaring an item -- a function,
2311 enumeration, structure, type, static, trait, implementation or module -- locally
2312 within a statement block is simply a way of restricting its scope to a narrow
2313 region containing all of its uses; it is otherwise identical in meaning to
2314 declaring the item outside the statement block.
2316 Note: there is no implicit capture of the function's dynamic environment when
2317 declaring a function-local item.
2319 #### Slot declarations
2321 ~~~~ {.notrust .ebnf .gram}
2322 let_decl : "let" pat [':' type ] ? [ init ] ? ';' ;
2323 init : [ '=' ] expr ;
2326 A _slot declaration_ introduces a new set of slots, given by a pattern.
2327 The pattern may be followed by a type annotation, and/or an initializer expression.
2328 When no type annotation is given, the compiler will infer the type,
2329 or signal an error if insufficient type information is available for definite inference.
2330 Any slots introduced by a slot declaration are visible from the point of declaration until the end of the enclosing block scope.
2332 ### Expression statements
2334 An _expression statement_ is one that evaluates an [expression](#expressions)
2335 and ignores its result.
2336 The type of an expression statement `e;` is always `()`, regardless of the type of `e`.
2337 As a rule, an expression statement's purpose is to trigger the effects of evaluating its expression.
2341 An expression may have two roles: it always produces a *value*, and it may have *effects*
2342 (otherwise known as "side effects").
2343 An expression *evaluates to* a value, and has effects during *evaluation*.
2344 Many expressions contain sub-expressions (operands).
2345 The meaning of each kind of expression dictates several things:
2346 * Whether or not to evaluate the sub-expressions when evaluating the expression
2347 * The order in which to evaluate the sub-expressions
2348 * How to combine the sub-expressions' values to obtain the value of the expression.
2350 In this way, the structure of expressions dictates the structure of execution.
2351 Blocks are just another kind of expression,
2352 so blocks, statements, expressions, and blocks again can recursively nest inside each other
2353 to an arbitrary depth.
2355 #### Lvalues, rvalues and temporaries
2357 Expressions are divided into two main categories: _lvalues_ and _rvalues_.
2358 Likewise within each expression, sub-expressions may occur in _lvalue context_ or _rvalue context_.
2359 The evaluation of an expression depends both on its own category and the context it occurs within.
2361 An lvalue is an expression that represents a memory location. These
2362 expressions are [paths](#path-expressions) (which refer to local
2363 variables, function and method arguments, or static variables),
2364 dereferences (`*expr`), [indexing expressions](#index-expressions)
2365 (`expr[expr]`), and [field references](#field-expressions) (`expr.f`).
2366 All other expressions are rvalues.
2368 The left operand of an [assignment](#assignment-expressions) or
2369 [compound-assignment](#compound-assignment-expressions) expression is an lvalue context,
2370 as is the single operand of a unary [borrow](#unary-operator-expressions).
2371 All other expression contexts are rvalue contexts.
2373 When an lvalue is evaluated in an _lvalue context_, it denotes a memory location;
2374 when evaluated in an _rvalue context_, it denotes the value held _in_ that memory location.
2376 When an rvalue is used in lvalue context, a temporary un-named lvalue is created and used instead.
2377 A temporary's lifetime equals the largest lifetime of any reference that points to it.
2379 #### Moved and copied types
2381 When a [local variable](#memory-slots) is used
2382 as an [rvalue](#lvalues-rvalues-and-temporaries)
2383 the variable will either be moved or copied, depending on its type.
2384 For types that contain [owning pointers](#pointer-types)
2385 or values that implement the special trait `Drop`,
2386 the variable is moved.
2387 All other types are copied.
2389 ### Literal expressions
2391 A _literal expression_ consists of one of the [literal](#literals)
2392 forms described earlier. It directly describes a number, character,
2393 string, boolean value, or the unit value.
2397 "hello"; // string type
2398 '5'; // character type
2402 ### Path expressions
2404 A [path](#paths) used as an expression context denotes either a local variable or an item.
2405 Path expressions are [lvalues](#lvalues-rvalues-and-temporaries).
2407 ### Tuple expressions
2409 Tuples are written by enclosing one or more comma-separated
2410 expressions in parentheses. They are used to create [tuple-typed](#tuple-types)
2419 ### Structure expressions
2421 ~~~~ {.notrust .ebnf .gram}
2422 struct_expr : expr_path '{' ident ':' expr
2423 [ ',' ident ':' expr ] *
2426 [ ',' expr ] * ')' |
2430 There are several forms of structure expressions.
2431 A _structure expression_ consists of the [path](#paths) of a [structure item](#structures),
2432 followed by a brace-enclosed list of one or more comma-separated name-value pairs,
2433 providing the field values of a new instance of the structure.
2434 A field name can be any identifier, and is separated from its value expression by a colon.
2435 The location denoted by a structure field is mutable if and only if the enclosing structure is mutable.
2437 A _tuple structure expression_ consists of the [path](#paths) of a [structure item](#structures),
2438 followed by a parenthesized list of one or more comma-separated expressions
2439 (in other words, the path of a structure item followed by a tuple expression).
2440 The structure item must be a tuple structure item.
2442 A _unit-like structure expression_ consists only of the [path](#paths) of a [structure item](#structures).
2444 The following are examples of structure expressions:
2447 # struct Point { x: f64, y: f64 }
2448 # struct TuplePoint(f64, f64);
2449 # mod game { pub struct User<'a> { pub name: &'a str, pub age: uint, pub score: uint } }
2450 # struct Cookie; fn some_fn<T>(t: T) {}
2451 Point {x: 10.0, y: 20.0};
2452 TuplePoint(10.0, 20.0);
2453 let u = game::User {name: "Joe", age: 35, score: 100_000};
2454 some_fn::<Cookie>(Cookie);
2457 A structure expression forms a new value of the named structure type.
2458 Note that for a given *unit-like* structure type, this will always be the same value.
2460 A structure expression can terminate with the syntax `..` followed by an expression to denote a functional update.
2461 The expression following `..` (the base) must have the same structure type as the new structure type being formed.
2462 The entire expression denotes the result of allocating a new structure
2463 (with the same type as the base expression)
2464 with the given values for the fields that were explicitly specified
2465 and the values in the base record for all other fields.
2468 # struct Point3d { x: int, y: int, z: int }
2469 let base = Point3d {x: 1, y: 2, z: 3};
2470 Point3d {y: 0, z: 10, .. base};
2473 ### Block expressions
2475 ~~~~ {.notrust .ebnf .gram}
2476 block_expr : '{' [ view_item ] *
2477 [ stmt ';' | item ] *
2481 A _block expression_ is similar to a module in terms of the declarations that
2482 are possible. Each block conceptually introduces a new namespace scope. View
2483 items can bring new names into scopes and declared items are in scope for only
2486 A block will execute each statement sequentially, and then execute the
2487 expression (if given). If the final expression is omitted, the type and return
2488 value of the block are `()`, but if it is provided, the type and return value
2489 of the block are that of the expression itself.
2491 ### Method-call expressions
2493 ~~~~ {.notrust .ebnf .gram}
2494 method_call_expr : expr '.' ident paren_expr_list ;
2497 A _method call_ consists of an expression followed by a single dot, an identifier, and a parenthesized expression-list.
2498 Method calls are resolved to methods on specific traits,
2499 either statically dispatching to a method if the exact `self`-type of the left-hand-side is known,
2500 or dynamically dispatching if the left-hand-side expression is an indirect [object type](#object-types).
2502 ### Field expressions
2504 ~~~~ {.notrust .ebnf .gram}
2505 field_expr : expr '.' ident ;
2508 A _field expression_ consists of an expression followed by a single dot and an identifier,
2509 when not immediately followed by a parenthesized expression-list (the latter is a [method call expression](#method-call-expressions)).
2510 A field expression denotes a field of a [structure](#structure-types).
2512 ~~~~ {.ignore .field}
2515 (Struct {a: 10, b: 20}).a;
2518 A field access on a record is an [lvalue](#lvalues-rvalues-and-temporaries) referring to the value of that field.
2519 When the field is mutable, it can be [assigned](#assignment-expressions) to.
2521 When the type of the expression to the left of the dot is a pointer to a record or structure,
2522 it is automatically dereferenced to make the field access possible.
2524 ### Vector expressions
2526 ~~~~ {.notrust .ebnf .gram}
2527 vec_expr : '[' "mut" ? vec_elems? ']' ;
2529 vec_elems : [expr [',' expr]*] | [expr ',' ".." expr] ;
2532 A [_vector_](#vector-types) _expression_ is written by enclosing zero or
2533 more comma-separated expressions of uniform type in square brackets.
2535 In the `[expr ',' ".." expr]` form, the expression after the `".."`
2536 must be a constant expression that can be evaluated at compile time, such
2537 as a [literal](#literals) or a [static item](#static-items).
2541 ["a", "b", "c", "d"];
2542 [0, ..128]; // vector with 128 zeros
2543 [0u8, 0u8, 0u8, 0u8];
2546 ### Index expressions
2548 ~~~~ {.notrust .ebnf .gram}
2549 idx_expr : expr '[' expr ']' ;
2552 [Vector](#vector-types)-typed expressions can be indexed by writing a
2553 square-bracket-enclosed expression (the index) after them. When the
2554 vector is mutable, the resulting [lvalue](#lvalues-rvalues-and-temporaries) can be assigned to.
2556 Indices are zero-based, and may be of any integral type. Vector access
2557 is bounds-checked at run-time. When the check fails, it will put the
2558 task in a _failing state_.
2562 # task::spawn(proc() {
2565 (["a", "b"])[10]; // fails
2570 ### Unary operator expressions
2572 Rust defines six symbolic unary operators.
2573 They are all written as prefix operators,
2574 before the expression they apply to.
2577 : Negation. May only be applied to numeric types.
2579 : Dereference. When applied to a [pointer](#pointer-types) it denotes the pointed-to location.
2580 For pointers to mutable locations, the resulting [lvalue](#lvalues-rvalues-and-temporaries) can be assigned to.
2581 On non-pointer types, it calls the `deref` method of the `std::ops::Deref` trait, or the
2582 `deref_mut` method of the `std::ops::DerefMut` trait (if implemented by the type and required
2583 for an outer expression that will or could mutate the dereference), and produces the
2584 result of dereferencing the `&` or `&mut` borrowed pointer returned from the overload method.
2587 : Logical negation. On the boolean type, this flips between `true` and
2588 `false`. On integer types, this inverts the individual bits in the
2589 two's complement representation of the value.
2591 : [Boxing](#pointer-types) operators. Allocate a box to hold the value they are applied to,
2592 and store the value in it. `~` creates an owned box.
2594 : Borrow operator. Returns a reference, pointing to its operand.
2595 The operand of a borrow is statically proven to outlive the resulting pointer.
2596 If the borrow-checker cannot prove this, it is a compilation error.
2598 ### Binary operator expressions
2600 ~~~~ {.notrust .ebnf .gram}
2601 binop_expr : expr binop expr ;
2604 Binary operators expressions are given in terms of
2605 [operator precedence](#operator-precedence).
2607 #### Arithmetic operators
2609 Binary arithmetic expressions are syntactic sugar for calls to built-in traits,
2610 defined in the `std::ops` module of the `std` library.
2611 This means that arithmetic operators can be overridden for user-defined types.
2612 The default meaning of the operators on standard types is given here.
2615 : Addition and vector/string concatenation.
2616 Calls the `add` method on the `std::ops::Add` trait.
2619 Calls the `sub` method on the `std::ops::Sub` trait.
2622 Calls the `mul` method on the `std::ops::Mul` trait.
2625 Calls the `div` method on the `std::ops::Div` trait.
2628 Calls the `rem` method on the `std::ops::Rem` trait.
2630 #### Bitwise operators
2632 Like the [arithmetic operators](#arithmetic-operators), bitwise operators
2633 are syntactic sugar for calls to methods of built-in traits.
2634 This means that bitwise operators can be overridden for user-defined types.
2635 The default meaning of the operators on standard types is given here.
2639 Calls the `bitand` method of the `std::ops::BitAnd` trait.
2642 Calls the `bitor` method of the `std::ops::BitOr` trait.
2645 Calls the `bitxor` method of the `std::ops::BitXor` trait.
2647 : Logical left shift.
2648 Calls the `shl` method of the `std::ops::Shl` trait.
2650 : Logical right shift.
2651 Calls the `shr` method of the `std::ops::Shr` trait.
2653 #### Lazy boolean operators
2655 The operators `||` and `&&` may be applied to operands of boolean type.
2656 The `||` operator denotes logical 'or', and the `&&` operator denotes logical 'and'.
2657 They differ from `|` and `&` in that the right-hand operand is only evaluated
2658 when the left-hand operand does not already determine the result of the expression.
2659 That is, `||` only evaluates its right-hand operand
2660 when the left-hand operand evaluates to `false`, and `&&` only when it evaluates to `true`.
2662 #### Comparison operators
2664 Comparison operators are, like the [arithmetic operators](#arithmetic-operators),
2665 and [bitwise operators](#bitwise-operators),
2666 syntactic sugar for calls to built-in traits.
2667 This means that comparison operators can be overridden for user-defined types.
2668 The default meaning of the operators on standard types is given here.
2672 Calls the `eq` method on the `std::cmp::Eq` trait.
2675 Calls the `ne` method on the `std::cmp::Eq` trait.
2678 Calls the `lt` method on the `std::cmp::Ord` trait.
2681 Calls the `gt` method on the `std::cmp::Ord` trait.
2683 : Less than or equal.
2684 Calls the `le` method on the `std::cmp::Ord` trait.
2686 : Greater than or equal.
2687 Calls the `ge` method on the `std::cmp::Ord` trait.
2689 #### Type cast expressions
2691 A type cast expression is denoted with the binary operator `as`.
2693 Executing an `as` expression casts the value on the left-hand side to the type
2694 on the right-hand side.
2696 A numeric value can be cast to any numeric type.
2697 A raw pointer value can be cast to or from any integral type or raw pointer type.
2698 Any other cast is unsupported and will fail to compile.
2700 An example of an `as` expression:
2703 # fn sum(v: &[f64]) -> f64 { 0.0 }
2704 # fn len(v: &[f64]) -> int { 0 }
2706 fn avg(v: &[f64]) -> f64 {
2707 let sum: f64 = sum(v);
2708 let sz: f64 = len(v) as f64;
2713 #### Assignment expressions
2715 An _assignment expression_ consists of an [lvalue](#lvalues-rvalues-and-temporaries) expression followed by an
2716 equals sign (`=`) and an [rvalue](#lvalues-rvalues-and-temporaries) expression.
2718 Evaluating an assignment expression [either copies or moves](#moved-and-copied-types) its right-hand operand to its left-hand operand.
2727 #### Compound assignment expressions
2729 The `+`, `-`, `*`, `/`, `%`, `&`, `|`, `^`, `<<`, and `>>`
2730 operators may be composed with the `=` operator. The expression `lval
2731 OP= val` is equivalent to `lval = lval OP val`. For example, `x = x +
2732 1` may be written as `x += 1`.
2734 Any such expression always has the [`unit`](#primitive-types) type.
2736 #### Operator precedence
2738 The precedence of Rust binary operators is ordered as follows, going
2739 from strong to weak:
2741 ~~~~ {.notrust .precedence}
2756 Operators at the same precedence level are evaluated left-to-right. [Unary operators](#unary-operator-expressions)
2757 have the same precedence level and it is stronger than any of the binary operators'.
2759 ### Grouped expressions
2761 An expression enclosed in parentheses evaluates to the result of the enclosed
2762 expression. Parentheses can be used to explicitly specify evaluation order
2763 within an expression.
2765 ~~~~ {.notrust .ebnf .gram}
2766 paren_expr : '(' expr ')' ;
2769 An example of a parenthesized expression:
2772 let x = (2 + 3) * 4;
2776 ### Call expressions
2778 ~~~~ {.notrust .ebnf .gram}
2779 expr_list : [ expr [ ',' expr ]* ] ? ;
2780 paren_expr_list : '(' expr_list ')' ;
2781 call_expr : expr paren_expr_list ;
2784 A _call expression_ invokes a function, providing zero or more input slots and
2785 an optional reference slot to serve as the function's output, bound to the
2786 `lval` on the right hand side of the call. If the function eventually returns,
2787 then the expression completes.
2789 Some examples of call expressions:
2792 # use std::from_str::FromStr;
2793 # fn add(x: int, y: int) -> int { 0 }
2795 let x: int = add(1, 2);
2796 let pi: Option<f32> = FromStr::from_str("3.14");
2799 ### Lambda expressions
2801 ~~~~ {.notrust .ebnf .gram}
2802 ident_list : [ ident [ ',' ident ]* ] ? ;
2803 lambda_expr : '|' ident_list '|' expr ;
2806 A _lambda expression_ (sometimes called an "anonymous function expression") defines a function and denotes it as a value,
2807 in a single expression.
2808 A lambda expression is a pipe-symbol-delimited (`|`) list of identifiers followed by an expression.
2810 A lambda expression denotes a function that maps a list of parameters (`ident_list`)
2811 onto the expression that follows the `ident_list`.
2812 The identifiers in the `ident_list` are the parameters to the function.
2813 These parameters' types need not be specified, as the compiler infers them from context.
2815 Lambda expressions are most useful when passing functions as arguments to other functions,
2816 as an abbreviation for defining and capturing a separate function.
2818 Significantly, lambda expressions _capture their environment_,
2819 which regular [function definitions](#functions) do not.
2820 The exact type of capture depends on the [function type](#function-types) inferred for the lambda expression.
2821 In the simplest and least-expensive form (analogous to a ```|| { }``` expression),
2822 the lambda expression captures its environment by reference,
2823 effectively borrowing pointers to all outer variables mentioned inside the function.
2824 Alternately, the compiler may infer that a lambda expression should copy or move values (depending on their type.)
2825 from the environment into the lambda expression's captured environment.
2827 In this example, we define a function `ten_times` that takes a higher-order function argument,
2828 and call it with a lambda expression as an argument.
2831 fn ten_times(f: |int|) {
2839 ten_times(|j| println!("hello, {}", j));
2844 ~~~~ {.notrust .ebnf .gram}
2845 while_expr : "while" expr '{' block '}' ;
2848 A `while` loop begins by evaluating the boolean loop conditional expression.
2849 If the loop conditional expression evaluates to `true`, the loop body block
2850 executes and control returns to the loop conditional expression. If the loop
2851 conditional expression evaluates to `false`, the `while` expression completes.
2866 The keyword `loop` in Rust appears both in _loop expressions_ and in _continue expressions_.
2867 A loop expression denotes an infinite loop;
2868 see [Continue expressions](#continue-expressions) for continue expressions.
2870 ~~~~ {.notrust .ebnf .gram}
2871 loop_expr : [ lifetime ':' ] "loop" '{' block '}';
2874 A `loop` expression may optionally have a _label_.
2875 If a label is present,
2876 then labeled `break` and `loop` expressions nested within this loop may exit out of this loop or return control to its head.
2877 See [Break expressions](#break-expressions).
2879 ### Break expressions
2881 ~~~~ {.notrust .ebnf .gram}
2882 break_expr : "break" [ lifetime ];
2885 A `break` expression has an optional `label`.
2886 If the label is absent, then executing a `break` expression immediately terminates the innermost loop enclosing it.
2887 It is only permitted in the body of a loop.
2888 If the label is present, then `break foo` terminates the loop with label `foo`,
2889 which need not be the innermost label enclosing the `break` expression,
2890 but must enclose it.
2892 ### Continue expressions
2894 ~~~~ {.notrust .ebnf .gram}
2895 continue_expr : "loop" [ lifetime ];
2898 A continue expression, written `loop`, also has an optional `label`.
2899 If the label is absent,
2900 then executing a `loop` expression immediately terminates the current iteration of the innermost loop enclosing it,
2901 returning control to the loop *head*.
2902 In the case of a `while` loop,
2903 the head is the conditional expression controlling the loop.
2904 In the case of a `for` loop, the head is the call-expression controlling the loop.
2905 If the label is present, then `loop foo` returns control to the head of the loop with label `foo`,
2906 which need not be the innermost label enclosing the `break` expression,
2907 but must enclose it.
2909 A `loop` expression is only permitted in the body of a loop.
2913 ~~~~ {.notrust .ebnf .gram}
2914 for_expr : "for" pat "in" expr '{' block '}' ;
2917 A `for` expression is a syntactic construct for looping over elements
2918 provided by an implementation of `std::iter::Iterator`.
2920 An example of a for loop over the contents of a vector:
2924 # fn bar(f: Foo) { }
2929 let v: &[Foo] = &[a, b, c];
2936 An example of a for loop over a series of integers:
2939 # fn bar(b:uint) { }
2940 for i in range(0u, 256) {
2947 ~~~~ {.notrust .ebnf .gram}
2948 if_expr : "if" expr '{' block '}'
2951 else_tail : "else" [ if_expr
2955 An `if` expression is a conditional branch in program control. The form of
2956 an `if` expression is a condition expression, followed by a consequent
2957 block, any number of `else if` conditions and blocks, and an optional
2958 trailing `else` block. The condition expressions must have type
2959 `bool`. If a condition expression evaluates to `true`, the
2960 consequent block is executed and any subsequent `else if` or `else`
2961 block is skipped. If a condition expression evaluates to `false`, the
2962 consequent block is skipped and any subsequent `else if` condition is
2963 evaluated. If all `if` and `else if` conditions evaluate to `false`
2964 then any `else` block is executed.
2966 ### Match expressions
2968 ~~~~ {.notrust .ebnf .gram}
2969 match_expr : "match" expr '{' match_arm [ '|' match_arm ] * '}' ;
2971 match_arm : match_pat "=>" [ expr "," | '{' block '}' ] ;
2973 match_pat : pat [ ".." pat ] ? [ "if" expr ] ;
2976 A `match` expression branches on a *pattern*. The exact form of matching that
2977 occurs depends on the pattern. Patterns consist of some combination of
2978 literals, destructured vectors or enum constructors, structures, records and
2979 tuples, variable binding specifications, wildcards (`..`), and placeholders
2980 (`_`). A `match` expression has a *head expression*, which is the value to
2981 compare to the patterns. The type of the patterns must equal the type of the
2984 In a pattern whose head expression has an `enum` type, a placeholder (`_`)
2985 stands for a *single* data field, whereas a wildcard `..` stands for *all* the
2986 fields of a particular variant. For example:
2989 enum List<X> { Nil, Cons(X, ~List<X>) }
2991 let x: List<int> = Cons(10, ~Cons(11, ~Nil));
2994 Cons(_, ~Nil) => fail!("singleton list"),
2996 Nil => fail!("empty list")
3000 The first pattern matches lists constructed by applying `Cons` to any head
3001 value, and a tail value of `~Nil`. The second pattern matches _any_ list
3002 constructed with `Cons`, ignoring the values of its arguments. The difference
3003 between `_` and `..` is that the pattern `C(_)` is only type-correct if `C` has
3004 exactly one argument, while the pattern `C(..)` is type-correct for any enum
3005 variant `C`, regardless of how many arguments `C` has.
3007 Used inside a vector pattern, `..` stands for any number of elements. This
3008 wildcard can be used at most once for a given vector, which implies that it
3009 cannot be used to specifically match elements that are at an unknown distance
3010 from both ends of a vector, like `[.., 42, ..]`. If followed by a variable name,
3011 it will bind the corresponding slice to the variable. Example:
3014 fn is_symmetric(list: &[uint]) -> bool {
3017 [x, ..inside, y] if x == y => is_symmetric(inside),
3023 let sym = &[0, 1, 4, 2, 4, 1, 0];
3024 let not_sym = &[0, 1, 7, 2, 4, 1, 0];
3025 assert!(is_symmetric(sym));
3026 assert!(!is_symmetric(not_sym));
3030 A `match` behaves differently depending on whether or not the head expression
3031 is an [lvalue or an rvalue](#lvalues-rvalues-and-temporaries).
3032 If the head expression is an rvalue, it is
3033 first evaluated into a temporary location, and the resulting value
3034 is sequentially compared to the patterns in the arms until a match
3035 is found. The first arm with a matching pattern is chosen as the branch target
3036 of the `match`, any variables bound by the pattern are assigned to local
3037 variables in the arm's block, and control enters the block.
3039 When the head expression is an lvalue, the match does not allocate a
3040 temporary location (however, a by-value binding may copy or move from
3041 the lvalue). When possible, it is preferable to match on lvalues, as the
3042 lifetime of these matches inherits the lifetime of the lvalue, rather
3043 than being restricted to the inside of the match.
3045 An example of a `match` expression:
3048 # fn process_pair(a: int, b: int) { }
3049 # fn process_ten() { }
3051 enum List<X> { Nil, Cons(X, ~List<X>) }
3053 let x: List<int> = Cons(10, ~Cons(11, ~Nil));
3056 Cons(a, ~Cons(b, _)) => {
3071 Patterns that bind variables
3072 default to binding to a copy or move of the matched value
3073 (depending on the matched value's type).
3074 This can be changed to bind to a reference by
3075 using the `ref` keyword,
3076 or to a mutable reference using `ref mut`.
3078 Subpatterns can also be bound to variables by the use of the syntax
3079 `variable @ pattern`.
3083 enum List { Nil, Cons(uint, ~List) }
3085 fn is_sorted(list: &List) -> bool {
3087 Nil | Cons(_, ~Nil) => true,
3088 Cons(x, ref r @ ~Cons(y, _)) => (x <= y) && is_sorted(*r)
3093 let a = Cons(6, ~Cons(7, ~Cons(42, ~Nil)));
3094 assert!(is_sorted(&a));
3099 Patterns can also dereference pointers by using the `&`,
3100 `~` or `@` symbols, as appropriate. For example, these two matches
3101 on `x: &int` are equivalent:
3105 let y = match *x { 0 => "zero", _ => "some" };
3106 let z = match x { &0 => "zero", _ => "some" };
3111 A pattern that's just an identifier, like `Nil` in the previous example,
3112 could either refer to an enum variant that's in scope, or bind a new variable.
3113 The compiler resolves this ambiguity by forbidding variable bindings that occur
3114 in `match` patterns from shadowing names of variants that are in scope.
3115 For example, wherever `List` is in scope,
3116 a `match` pattern would not be able to bind `Nil` as a new name.
3117 The compiler interprets a variable pattern `x` as a binding _only_ if there is
3118 no variant named `x` in scope.
3119 A convention you can use to avoid conflicts is simply to name variants with
3120 upper-case letters, and local variables with lower-case letters.
3122 Multiple match patterns may be joined with the `|` operator.
3123 A range of values may be specified with `..`.
3129 let message = match x {
3130 0 | 1 => "not many",
3136 Range patterns only work on scalar types
3137 (like integers and characters; not like vectors and structs, which have sub-components).
3138 A range pattern may not be a sub-range of another range pattern inside the same `match`.
3140 Finally, match patterns can accept *pattern guards* to further refine the
3141 criteria for matching a case. Pattern guards appear after the pattern and
3142 consist of a bool-typed expression following the `if` keyword. A pattern
3143 guard may refer to the variables bound within the pattern they follow.
3146 # let maybe_digit = Some(0);
3147 # fn process_digit(i: int) { }
3148 # fn process_other(i: int) { }
3150 let message = match maybe_digit {
3151 Some(x) if x < 10 => process_digit(x),
3152 Some(x) => process_other(x),
3157 ### Return expressions
3159 ~~~~ {.notrust .ebnf .gram}
3160 return_expr : "return" expr ? ;
3163 Return expressions are denoted with the keyword `return`. Evaluating a `return`
3164 expression moves its argument into the output slot of the current
3165 function, destroys the current function activation frame, and transfers
3166 control to the caller frame.
3168 An example of a `return` expression:
3171 fn max(a: int, b: int) -> int {
3183 Every slot, item and value in a Rust program has a type. The _type_ of a *value*
3184 defines the interpretation of the memory holding it.
3186 Built-in types and type-constructors are tightly integrated into the language,
3187 in nontrivial ways that are not possible to emulate in user-defined
3188 types. User-defined types have limited capabilities.
3192 The primitive types are the following:
3194 * The "unit" type `()`, having the single "unit" value `()` (occasionally called "nil").
3195 ^[The "unit" value `()` is *not* a sentinel "null pointer" value for reference slots; the "unit" type is the implicit return type from functions otherwise lacking a return type, and can be used in other contexts (such as message-sending or type-parametric code) as a zero-size type.]
3196 * The boolean type `bool` with values `true` and `false`.
3197 * The machine types.
3198 * The machine-dependent integer and floating-point types.
3202 The machine types are the following:
3204 * The unsigned word types `u8`, `u16`, `u32` and `u64`, with values drawn from
3205 the integer intervals $[0, 2^8 - 1]$, $[0, 2^{16} - 1]$, $[0, 2^{32} - 1]$ and
3206 $[0, 2^{64} - 1]$ respectively.
3208 * The signed two's complement word types `i8`, `i16`, `i32` and `i64`, with
3209 values drawn from the integer intervals $[-(2^7), 2^7 - 1]$,
3210 $[-(2^{15}), 2^{15} - 1]$, $[-(2^{31}), 2^{31} - 1]$, $[-(2^{63}), 2^{63} - 1]$
3213 * The IEEE 754-2008 `binary32` and `binary64` floating-point types: `f32` and
3214 `f64`, respectively.
3216 #### Machine-dependent integer types
3218 The Rust type `uint`^[A Rust `uint` is analogous to a C99 `uintptr_t`.] is an
3219 unsigned integer type with target-machine-dependent size. Its size, in
3220 bits, is equal to the number of bits required to hold any memory address on
3223 The Rust type `int`^[A Rust `int` is analogous to a C99 `intptr_t`.] is a
3224 two's complement signed integer type with target-machine-dependent size. Its
3225 size, in bits, is equal to the size of the rust type `uint` on the same target
3230 The types `char` and `str` hold textual data.
3232 A value of type `char` is a [Unicode scalar value](
3233 http://www.unicode.org/glossary/#unicode_scalar_value)
3234 (ie. a code point that is not a surrogate),
3235 represented as a 32-bit unsigned word in the 0x0000 to 0xD7FF
3236 or 0xE000 to 0x10FFFF range.
3237 A `[char]` vector is effectively an UCS-4 / UTF-32 string.
3239 A value of type `str` is a Unicode string,
3240 represented as a vector of 8-bit unsigned bytes holding a sequence of UTF-8 codepoints.
3241 Since `str` is of unknown size, it is not a _first class_ type,
3242 but can only be instantiated through a pointer type,
3243 such as `&str` or `~str`.
3247 The tuple type-constructor forms a new heterogeneous product of values similar
3248 to the record type-constructor. The differences are as follows:
3250 * tuple elements cannot be mutable, unlike record fields
3251 * tuple elements are not named and can be accessed only by pattern-matching
3253 Tuple types and values are denoted by listing the types or values of their
3254 elements, respectively, in a parenthesized, comma-separated
3257 The members of a tuple are laid out in memory contiguously, like a record, in
3258 order specified by the tuple type.
3260 An example of a tuple type and its use:
3263 type Pair<'a> = (int,&'a str);
3264 let p: Pair<'static> = (10,"hello");
3266 assert!(b != "world");
3271 The vector type constructor represents a homogeneous array of values of a given type.
3272 A vector has a fixed size.
3273 (Operations like `vec.push` operate solely on owned vectors.)
3274 A vector type can be annotated with a _definite_ size, such as `[int, ..10]`.
3275 Such a definite-sized vector type is a first-class type, since its size is known statically.
3276 A vector without such a size is said to be of _indefinite_ size,
3277 and is therefore not a _first-class_ type.
3278 An indefinite-size vector can only be instantiated through a pointer type,
3279 such as `&[T]` or `~[T]`.
3280 The kind of a vector type depends on the kind of its element type,
3281 as with other simple structural types.
3283 Expressions producing vectors of definite size cannot be evaluated in a
3284 context expecting a vector of indefinite size; one must copy the
3285 definite-sized vector contents into a distinct vector of indefinite size.
3287 An example of a vector type and its use:
3290 let v: &[int] = &[7, 5, 3];
3295 All in-bounds elements of a vector are always initialized,
3296 and access to a vector is always bounds-checked.
3300 A `struct` *type* is a heterogeneous product of other types, called the *fields* of the type.
3301 ^[`struct` types are analogous `struct` types in C,
3302 the *record* types of the ML family,
3303 or the *structure* types of the Lisp family.]
3305 New instances of a `struct` can be constructed with a [struct expression](#structure-expressions).
3307 The memory order of fields in a `struct` is given by the item defining it.
3308 Fields may be given in any order in a corresponding struct *expression*;
3309 the resulting `struct` value will always be laid out in memory in the order specified by the corresponding *item*.
3311 The fields of a `struct` may be qualified by [visibility modifiers](#re-exporting-and-visibility),
3312 to restrict access to implementation-private data in a structure.
3314 A _tuple struct_ type is just like a structure type, except that the fields are anonymous.
3316 A _unit-like struct_ type is like a structure type, except that it has no fields.
3317 The one value constructed by the associated [structure expression](#structure-expressions)
3318 is the only value that inhabits such a type.
3320 ### Enumerated types
3322 An *enumerated type* is a nominal, heterogeneous disjoint union type,
3323 denoted by the name of an [`enum` item](#enumerations).
3324 ^[The `enum` type is analogous to a `data` constructor declaration in ML,
3325 or a *pick ADT* in Limbo.]
3327 An [`enum` item](#enumerations) declares both the type and a number of *variant constructors*,
3328 each of which is independently named and takes an optional tuple of arguments.
3330 New instances of an `enum` can be constructed by calling one of the variant constructors,
3331 in a [call expression](#call-expressions).
3333 Any `enum` value consumes as much memory as the largest variant constructor for its corresponding `enum` type.
3335 Enum types cannot be denoted *structurally* as types,
3336 but must be denoted by named reference to an [`enum` item](#enumerations).
3340 Nominal types -- [enumerations](#enumerated-types) and [structures](#structure-types) -- may be recursive.
3341 That is, each `enum` constructor or `struct` field may refer, directly or indirectly, to the enclosing `enum` or `struct` type itself.
3342 Such recursion has restrictions:
3344 * Recursive types must include a nominal type in the recursion
3345 (not mere [type definitions](#type-definitions),
3346 or other structural types such as [vectors](#vector-types) or [tuples](#tuple-types)).
3347 * A recursive `enum` item must have at least one non-recursive constructor
3348 (in order to give the recursion a basis case).
3349 * The size of a recursive type must be finite;
3350 in other words the recursive fields of the type must be [pointer types](#pointer-types).
3351 * Recursive type definitions can cross module boundaries, but not module *visibility* boundaries,
3352 or crate boundaries (in order to simplify the module system and type checker).
3354 An example of a *recursive* type and its use:
3362 let a: List<int> = Cons(7, ~Cons(13, ~Nil));
3367 All pointers in Rust are explicit first-class values.
3368 They can be copied, stored into data structures, and returned from functions.
3369 There are four varieties of pointer in Rust:
3371 Owning pointers (`~`)
3372 : These point to owned heap allocations (or "boxes") in the shared, inter-task heap.
3373 Each owned box has a single owning pointer; pointer and pointee retain a 1:1 relationship at all times.
3374 Owning pointers are written `~content`,
3375 for example `~int` means an owning pointer to an owned box containing an integer.
3376 Copying an owned box is a "deep" operation:
3377 it involves allocating a new owned box and copying the contents of the old box into the new box.
3378 Releasing an owning pointer immediately releases its corresponding owned box.
3381 : These point to memory _owned by some other value_.
3382 References arise by (automatic) conversion from owning pointers, managed pointers,
3383 or by applying the borrowing operator `&` to some other value,
3384 including [lvalues, rvalues or temporaries](#lvalues-rvalues-and-temporaries).
3385 References are written `&content`, or in some cases `&'f content` for some lifetime-variable `f`,
3386 for example `&int` means a reference to an integer.
3387 Copying a reference is a "shallow" operation:
3388 it involves only copying the pointer itself.
3389 Releasing a reference typically has no effect on the value it points to,
3390 with the exception of temporary values,
3391 which are released when the last reference to them is released.
3394 : Raw pointers are pointers without safety or liveness guarantees.
3395 Raw pointers are written `*content`,
3396 for example `*int` means a raw pointer to an integer.
3397 Copying or dropping a raw pointer has no effect on the lifecycle of any other value.
3398 Dereferencing a raw pointer or converting it to any other pointer type is an [`unsafe` operation](#unsafe-functions).
3399 Raw pointers are generally discouraged in Rust code;
3400 they exist to support interoperability with foreign code,
3401 and writing performance-critical or low-level functions.
3405 The function type constructor `fn` forms new function types.
3406 A function type consists of a possibly-empty set of function-type modifiers
3407 (such as `unsafe` or `extern`), a sequence of input types and an output type.
3409 An example of a `fn` type:
3412 fn add(x: int, y: int) -> int {
3416 let mut x = add(5,7);
3418 type Binop<'a> = 'a |int,int| -> int;
3419 let bo: Binop = add;
3425 ~~~~ {.notrust .ebnf .notation}
3426 closure_type := [ 'unsafe' ] [ '<' lifetime-list '>' ] '|' arg-list '|'
3427 [ ':' bound-list ] [ '->' type ]
3428 procedure_type := 'proc' [ '<' lifetime-list '>' ] '(' arg-list ')'
3429 [ ':' bound-list ] [ '->' type ]
3430 lifetime-list := lifetime | lifetime ',' lifetime-list
3431 arg-list := ident ':' type | ident ':' type ',' arg-list
3432 bound-list := bound | bound '+' bound-list
3433 bound := path | lifetime
3436 The type of a closure mapping an input of type `A` to an output of type `B` is
3437 `|A| -> B`. A closure with no arguments or return values has type `||`.
3438 Similarly, a procedure mapping `A` to `B` is `proc(A) -> B` and a no-argument
3439 and no-return value closure has type `proc()`.
3441 An example of creating and calling a closure:
3444 let captured_var = 10;
3446 let closure_no_args = || println!("captured_var={}", captured_var);
3448 let closure_args = |arg: int| -> int {
3449 println!("captured_var={}, arg={}", captured_var, arg);
3450 arg // Note lack of semicolon after 'arg'
3453 fn call_closure(c1: ||, c2: |int| -> int) {
3458 call_closure(closure_no_args, closure_args);
3462 Unlike closures, procedures may only be invoked once, but own their
3463 environment, and are allowed to move out of their environment. Procedures are
3464 allocated on the heap (unlike closures). An example of creating and calling a
3468 let string = ~"Hello";
3470 // Creates a new procedure, passing it to the `spawn` function.
3472 println!("{} world!", string);
3475 // the variable `string` has been moved into the previous procedure, so it is
3476 // no longer usable.
3479 // Create an invoke a procedure. Note that the procedure is *moved* when
3480 // invoked, so it cannot be invoked again.
3481 let f = proc(n: int) { n + 22 };
3482 println!("answer: {}", f(20));
3488 Every trait item (see [traits](#traits)) defines a type with the same name as the trait.
3489 This type is called the _object type_ of the trait.
3490 Object types permit "late binding" of methods, dispatched using _virtual method tables_ ("vtables").
3491 Whereas most calls to trait methods are "early bound" (statically resolved) to specific implementations at compile time,
3492 a call to a method on an object type is only resolved to a vtable entry at compile time.
3493 The actual implementation for each vtable entry can vary on an object-by-object basis.
3495 Given a pointer-typed expression `E` of type `&T` or `~T`, where `T` implements trait `R`,
3496 casting `E` to the corresponding pointer type `&R` or `~R` results in a value of the _object type_ `R`.
3497 This result is represented as a pair of pointers:
3498 the vtable pointer for the `T` implementation of `R`, and the pointer value of `E`.
3500 An example of an object type:
3504 fn to_string(&self) -> ~str;
3507 impl Printable for int {
3508 fn to_string(&self) -> ~str { self.to_str() }
3511 fn print(a: ~Printable) {
3512 println!("{}", a.to_string());
3516 print(~10 as ~Printable);
3520 In this example, the trait `Printable` occurs as an object type in both the type signature of `print`,
3521 and the cast expression in `main`.
3525 Within the body of an item that has type parameter declarations, the names of its type parameters are types:
3528 fn map<A: Clone, B: Clone>(f: |A| -> B, xs: &[A]) -> ~[B] {
3532 let first: B = f(xs[0].clone());
3533 let rest: ~[B] = map(f, xs.slice(1, xs.len()));
3534 return ~[first] + rest;
3538 Here, `first` has type `B`, referring to `map`'s `B` type parameter;
3539 and `rest` has type `~[B]`, a vector type with element type `B`.
3543 The special type `self` has a meaning within methods inside an
3544 impl item. It refers to the type of the implicit `self` argument. For
3549 fn make_string(&self) -> ~str;
3552 impl Printable for ~str {
3553 fn make_string(&self) -> ~str {
3559 `self` refers to the value of type `~str` that is the receiver for a
3560 call to the method `make_string`.
3564 Types in Rust are categorized into kinds, based on various properties of the components of the type.
3568 : Types of this kind can be safely sent between tasks.
3569 This kind includes scalars, owning pointers, owned closures, and
3570 structural types containing only other owned types.
3571 All `Send` types are `'static`.
3573 : Types of this kind consist of "Plain Old Data"
3574 which can be copied by simply moving bits.
3575 All values of this kind can be implicitly copied.
3576 This kind includes scalars and immutable references,
3577 as well as structural types containing other `Copy` types.
3579 : Types of this kind do not contain any references (except for
3580 references with the `static` lifetime, which are allowed).
3581 This can be a useful guarantee for code
3582 that breaks borrowing assumptions
3583 using [`unsafe` operations](#unsafe-functions).
3585 : This is not strictly a kind,
3586 but its presence interacts with kinds:
3587 the `Drop` trait provides a single method `drop`
3588 that takes no parameters,
3589 and is run when values of the type are dropped.
3590 Such a method is called a "destructor",
3591 and are always executed in "top-down" order:
3592 a value is completely destroyed
3593 before any of the values it owns run their destructors.
3594 Only `Send` types can implement `Drop`.
3597 : Types with destructors, closure environments,
3598 and various other _non-first-class_ types,
3599 are not copyable at all.
3600 Such types can usually only be accessed through pointers,
3601 or in some cases, moved between mutable locations.
3603 Kinds can be supplied as _bounds_ on type parameters, like traits,
3604 in which case the parameter is constrained to types satisfying that kind.
3606 By default, type parameters do not carry any assumed kind-bounds at all.
3607 When instantiating a type parameter,
3608 the kind bounds on the parameter are checked
3609 to be the same or narrower than the kind
3610 of the type that it is instantiated with.
3612 Sending operations are not part of the Rust language,
3613 but are implemented in the library.
3614 Generic functions that send values
3615 bound the kind of these values to sendable.
3617 # Memory and concurrency models
3619 Rust has a memory model centered around concurrently-executing _tasks_. Thus
3620 its memory model and its concurrency model are best discussed simultaneously,
3621 as parts of each only make sense when considered from the perspective of the
3624 When reading about the memory model, keep in mind that it is partitioned in
3625 order to support tasks; and when reading about tasks, keep in mind that their
3626 isolation and communication mechanisms are only possible due to the ownership
3627 and lifetime semantics of the memory model.
3631 A Rust program's memory consists of a static set of *items*, a set of
3632 [tasks](#tasks) each with its own *stack*, and a *heap*. Immutable portions of
3633 the heap may be shared between tasks, mutable portions may not.
3635 Allocations in the stack consist of *slots*, and allocations in the heap
3638 ### Memory allocation and lifetime
3640 The _items_ of a program are those functions, modules and types
3641 that have their value calculated at compile-time and stored uniquely in the
3642 memory image of the rust process. Items are neither dynamically allocated nor
3645 A task's _stack_ consists of activation frames automatically allocated on
3646 entry to each function as the task executes. A stack allocation is reclaimed
3647 when control leaves the frame containing it.
3649 The _heap_ is a general term that describes two separate sets of boxes:
3650 managed boxes -- which may be subject to garbage collection -- and owned
3651 boxes. The lifetime of an allocation in the heap depends on the lifetime of
3652 the box values pointing to it. Since box values may themselves be passed in
3653 and out of frames, or stored in the heap, heap allocations may outlive the
3654 frame they are allocated within.
3656 ### Memory ownership
3658 A task owns all memory it can *safely* reach through local variables,
3659 as well as managed, owned boxes and references.
3661 When a task sends a value that has the `Send` trait to another task,
3662 it loses ownership of the value sent and can no longer refer to it.
3663 This is statically guaranteed by the combined use of "move semantics",
3664 and the compiler-checked _meaning_ of the `Send` trait:
3665 it is only instantiated for (transitively) sendable kinds of data constructor and pointers,
3666 never including managed boxes or references.
3668 When a stack frame is exited, its local allocations are all released, and its
3669 references to boxes (both managed and owned) are dropped.
3671 A managed box may (in the case of a recursive, mutable managed type) be cyclic;
3672 in this case the release of memory inside the managed structure may be deferred
3673 until task-local garbage collection can reclaim it. Code can ensure no such
3674 delayed deallocation occurs by restricting itself to owned boxes and similar
3675 unmanaged kinds of data.
3677 When a task finishes, its stack is necessarily empty and it therefore has no
3678 references to any boxes; the remainder of its heap is immediately freed.
3682 A task's stack contains slots.
3684 A _slot_ is a component of a stack frame, either a function parameter,
3685 a [temporary](#lvalues-rvalues-and-temporaries), or a local variable.
3687 A _local variable_ (or *stack-local* allocation) holds a value directly,
3688 allocated within the stack's memory. The value is a part of the stack frame.
3690 Local variables are immutable unless declared otherwise like: `let mut x = ...`.
3692 Function parameters are immutable unless declared with `mut`. The
3693 `mut` keyword applies only to the following parameter (so `|mut x, y|`
3694 and `fn f(mut x: ~int, y: ~int)` declare one mutable variable `x` and
3695 one immutable variable `y`).
3697 Methods that take either `self` or `~self` can optionally place them in a
3698 mutable slot by prefixing them with `mut` (similar to regular arguments):
3702 fn change(mut self) -> Self;
3703 fn modify(mut ~self) -> ~Self;
3707 Local variables are not initialized when allocated; the entire frame worth of
3708 local variables are allocated at once, on frame-entry, in an uninitialized
3709 state. Subsequent statements within a function may or may not initialize the
3710 local variables. Local variables can be used only after they have been
3711 initialized; this is enforced by the compiler.
3715 An _owned box_ is a reference to a heap allocation holding another value, which is constructed
3716 by the prefix *tilde* sigil `~`
3718 An example of an owned box type and value:
3724 Owned box values exist in 1:1 correspondence with their heap allocation
3725 copying an owned box value makes a shallow copy of the pointer
3726 Rust will consider a shallow copy of an owned box to move ownership of the value. After a value has been moved, the source location cannot be used unless it is reinitialized.
3731 // attempting to use `x` will result in an error here
3738 An executing Rust program consists of a tree of tasks.
3739 A Rust _task_ consists of an entry function, a stack,
3740 a set of outgoing communication channels and incoming communication ports,
3741 and ownership of some portion of the heap of a single operating-system process.
3742 (We expect that many programs will not use channels and ports directly,
3743 but will instead use higher-level abstractions provided in standard libraries,
3746 Multiple Rust tasks may coexist in a single operating-system process.
3747 The runtime scheduler maps tasks to a certain number of operating-system threads.
3748 By default, the scheduler chooses the number of threads based on
3749 the number of concurrent physical CPUs detected at startup.
3750 It's also possible to override this choice at runtime.
3751 When the number of tasks exceeds the number of threads -- which is likely --
3752 the scheduler multiplexes the tasks onto threads.^[
3753 This is an M:N scheduler,
3754 which is known to give suboptimal results for CPU-bound concurrency problems.
3755 In such cases, running with the same number of threads and tasks can yield better results.
3756 Rust has M:N scheduling in order to support very large numbers of tasks
3757 in contexts where threads are too resource-intensive to use in large number.
3758 The cost of threads varies substantially per operating system, and is sometimes quite low,
3759 so this flexibility is not always worth exploiting.]
3761 ### Communication between tasks
3763 Rust tasks are isolated and generally unable to interfere with one another's memory directly,
3764 except through [`unsafe` code](#unsafe-functions).
3765 All contact between tasks is mediated by safe forms of ownership transfer,
3766 and data races on memory are prohibited by the type system.
3768 Inter-task communication and co-ordination facilities are provided in the standard library.
3771 - synchronous and asynchronous communication channels with various communication topologies
3772 - read-only and read-write shared variables with various safe mutual exclusion patterns
3773 - simple locks and semaphores
3775 When such facilities carry values, the values are restricted to the [`Send` type-kind](#type-kinds).
3776 Restricting communication interfaces to this kind ensures that no references or managed pointers move between tasks.
3777 Thus access to an entire data structure can be mediated through its owning "root" value;
3778 no further locking or copying is required to avoid data races within the substructure of such a value.
3782 The _lifecycle_ of a task consists of a finite set of states and events
3783 that cause transitions between the states. The lifecycle states of a task are:
3790 A task begins its lifecycle -- once it has been spawned -- in the *running*
3791 state. In this state it executes the statements of its entry function, and any
3792 functions called by the entry function.
3794 A task may transition from the *running* state to the *blocked*
3795 state any time it makes a blocking communication call. When the
3796 call can be completed -- when a message arrives at a sender, or a
3797 buffer opens to receive a message -- then the blocked task will
3798 unblock and transition back to *running*.
3800 A task may transition to the *failing* state at any time, due being
3801 killed by some external event or internally, from the evaluation of a
3802 `fail!()` macro. Once *failing*, a task unwinds its stack and
3803 transitions to the *dead* state. Unwinding the stack of a task is done by
3804 the task itself, on its own control stack. If a value with a destructor is
3805 freed during unwinding, the code for the destructor is run, also on the task's
3806 control stack. Running the destructor code causes a temporary transition to a
3807 *running* state, and allows the destructor code to cause any subsequent
3808 state transitions. The original task of unwinding and failing thereby may
3809 suspend temporarily, and may involve (recursive) unwinding of the stack of a
3810 failed destructor. Nonetheless, the outermost unwinding activity will continue
3811 until the stack is unwound and the task transitions to the *dead*
3812 state. There is no way to "recover" from task failure. Once a task has
3813 temporarily suspended its unwinding in the *failing* state, failure
3814 occurring from within this destructor results in *hard* failure.
3815 A hard failure currently results in the process aborting.
3817 A task in the *dead* state cannot transition to other states; it exists
3818 only to have its termination status inspected by other tasks, and/or to await
3819 reclamation when the last reference to it drops.
3823 The currently scheduled task is given a finite *time slice* in which to
3824 execute, after which it is *descheduled* at a loop-edge or similar
3825 preemption point, and another task within is scheduled, pseudo-randomly.
3827 An executing task can yield control at any time, by making a library call to
3828 `std::task::yield`, which deschedules it immediately. Entering any other
3829 non-executing state (blocked, dead) similarly deschedules the task.
3831 # Runtime services, linkage and debugging
3833 The Rust _runtime_ is a relatively compact collection of C++ and Rust code
3834 that provides fundamental services and datatypes to all Rust tasks at
3835 run-time. It is smaller and simpler than many modern language runtimes. It is
3836 tightly integrated into the language's execution model of memory, tasks,
3837 communication and logging.
3839 > **Note:** The runtime library will merge with the `std` library in future versions of Rust.
3841 ### Memory allocation
3843 The runtime memory-management system is based on a _service-provider interface_,
3844 through which the runtime requests blocks of memory from its environment
3845 and releases them back to its environment when they are no longer needed.
3846 The default implementation of the service-provider interface
3847 consists of the C runtime functions `malloc` and `free`.
3849 The runtime memory-management system, in turn, supplies Rust tasks with
3850 facilities for allocating releasing stacks, as well as allocating and freeing
3855 The runtime provides C and Rust code to assist with various built-in types,
3856 such as vectors, strings, and the low level communication system (ports,
3859 Support for other built-in types such as simple types, tuples, records, and
3860 enums is open-coded by the Rust compiler.
3862 ### Task scheduling and communication
3864 The runtime provides code to manage inter-task communication. This includes
3865 the system of task-lifecycle state transitions depending on the contents of
3866 queues, as well as code to copy values between queues and their recipients and
3867 to serialize values for transmission over operating-system inter-process
3868 communication facilities.
3872 The Rust compiler supports various methods to link crates together both
3873 statically and dynamically. This section will explore the various methods to
3874 link Rust crates together, and more information about native libraries can be
3875 found in the [ffi tutorial][ffi].
3877 In one session of compilation, the compiler can generate multiple artifacts
3878 through the usage of command line flags and the `crate_type` attribute.
3880 * `--crate-type=bin`, `#[crate_type = "bin"]` - A runnable executable will be
3881 produced. This requires that there is a `main` function in the crate which
3882 will be run when the program begins executing. This will link in all Rust and
3883 native dependencies, producing a distributable binary.
3885 * `--crate-type=lib`, `#[crate_type = "lib"]` - A Rust library will be produced.
3886 This is an ambiguous concept as to what exactly is produced because a library
3887 can manifest itself in several forms. The purpose of this generic `lib` option
3888 is to generate the "compiler recommended" style of library. The output library
3889 will always be usable by rustc, but the actual type of library may change from
3890 time-to-time. The remaining output types are all different flavors of
3891 libraries, and the `lib` type can be seen as an alias for one of them (but the
3892 actual one is compiler-defined).
3894 * `--crate-type=dylib`, `#[crate_type = "dylib"]` - A dynamic Rust library will
3895 be produced. This is different from the `lib` output type in that this forces
3896 dynamic library generation. The resulting dynamic library can be used as a
3897 dependency for other libraries and/or executables. This output type will
3898 create `*.so` files on linux, `*.dylib` files on osx, and `*.dll` files on
3901 * `--crate-type=staticlib`, `#[crate_type = "staticlib"]` - A static system
3902 library will be produced. This is different from other library outputs in that
3903 the Rust compiler will never attempt to link to `staticlib` outputs. The
3904 purpose of this output type is to create a static library containing all of
3905 the local crate's code along with all upstream dependencies. The static
3906 library is actually a `*.a` archive on linux and osx and a `*.lib` file on
3907 windows. This format is recommended for use in situtations such as linking
3908 Rust code into an existing non-Rust application because it will not have
3909 dynamic dependencies on other Rust code.
3911 * `--crate-type=rlib`, `#[crate_type = "rlib"]` - A "Rust library" file will be
3912 produced. This is used as an intermediate artifact and can be thought of as a
3913 "static Rust library". These `rlib` files, unlike `staticlib` files, are
3914 interpreted by the Rust compiler in future linkage. This essentially means
3915 that `rustc` will look for metadata in `rlib` files like it looks for metadata
3916 in dynamic libraries. This form of output is used to produce statically linked
3917 executables as well as `staticlib` outputs.
3919 Note that these outputs are stackable in the sense that if multiple are
3920 specified, then the compiler will produce each form of output at once without
3921 having to recompile.
3923 With all these different kinds of outputs, if crate A depends on crate B, then
3924 the compiler could find B in various different forms throughout the system. The
3925 only forms looked for by the compiler, however, are the `rlib` format and the
3926 dynamic library format. With these two options for a dependent library, the
3927 compiler must at some point make a choice between these two formats. With this
3928 in mind, the compiler follows these rules when determining what format of
3929 dependencies will be used:
3931 1. If a dynamic library is being produced, then it is required for all upstream
3932 Rust dependencies to also be dynamic. This is a limitation of the current
3933 implementation of the linkage model. The reason behind this limitation is to
3934 prevent multiple copies of the same upstream library from showing up, and in
3935 the future it is planned to support a mixture of dynamic and static linking.
3937 When producing a dynamic library, the compiler will generate an error if an
3938 upstream dependency could not be found, and also if an upstream dependency
3939 could only be found in an `rlib` format. Remember that `staticlib` formats
3940 are always ignored by `rustc` for crate-linking purposes.
3942 2. If a static library is being produced, all upstream dependecies are
3943 required to be available in `rlib` formats. This requirement stems from the
3944 same reasons that a dynamic library must have all dynamic dependencies.
3946 Note that it is impossible to link in native dynamic dependencies to a static
3947 library, and in this case warnings will be printed about all unlinked native
3948 dynamic dependencies.
3950 3. If an `rlib` file is being produced, then there are no restrictions on what
3951 format the upstream dependencies are available in. It is simply required that
3952 all upstream dependencies be available for reading metadata from.
3954 The reason for this is that `rlib` files do not contain any of their upstream
3955 dependencies. It wouldn't be very efficient for all `rlib` files to contain a
3956 copy of `libstd.rlib`!
3958 4. If an executable is being produced, then things get a little interesting. As
3959 with the above limitations in dynamic and static libraries, it is required
3960 for all upstream dependencies to be in the same format. The next question is
3961 whether to prefer a dynamic or a static format. The compiler currently favors
3962 static linking over dynamic linking, but this can be inverted with the `-C
3963 prefer-dynamic` flag to the compiler.
3965 What this means is that first the compiler will attempt to find all upstream
3966 dependencies as `rlib` files, and if successful, it will create a statically
3967 linked executable. If an upstream dependency is missing as an `rlib` file,
3968 then the compiler will force all dependencies to be dynamic and will generate
3969 errors if dynamic versions could not be found.
3971 In general, `--crate-type=bin` or `--crate-type=lib` should be sufficient for
3972 all compilation needs, and the other options are just available if more
3973 fine-grained control is desired over the output format of a Rust crate.
3977 The runtime contains a system for directing [logging
3978 expressions](#logging-expressions) to a logging console and/or internal logging
3979 buffers. Logging can be enabled per module.
3981 Logging output is enabled by setting the `RUST_LOG` environment
3982 variable. `RUST_LOG` accepts a logging specification made up of a
3983 comma-separated list of paths, with optional log levels. For each
3984 module containing log expressions, if `RUST_LOG` contains the path to
3985 that module or a parent of that module, then logs of the appropriate
3986 level will be output to the console.
3988 The path to a module consists of the crate name, any parent modules,
3989 then the module itself, all separated by double colons (`::`). The
3990 optional log level can be appended to the module path with an equals
3991 sign (`=`) followed by the log level, from 1 to 4, inclusive. Level 1
3992 is the error level, 2 is warning, 3 info, and 4 debug. You can also
3993 use the symbolic constants `error`, `warn`, `info`, and `debug`. Any
3994 logs less than or equal to the specified level will be output. If not
3995 specified then log level 4 is assumed. Debug messages can be omitted
3996 by passing `--cfg ndebug` to `rustc`.
3998 As an example, to see all the logs generated by the compiler, you would set
3999 `RUST_LOG` to `rustc`, which is the crate name (as specified in its `crate_id`
4000 [attribute](#attributes)). To narrow down the logs to just crate resolution,
4001 you would set it to `rustc::metadata::creader`. To see just error logging
4004 Note that when compiling source files that don't specify a
4005 crate name the crate is given a default name that matches the source file,
4006 with the extension removed. In that case, to turn on logging for a program
4007 compiled from, e.g. `helloworld.rs`, `RUST_LOG` should be set to `helloworld`.
4009 As a convenience, the logging spec can also be set to a special pseudo-crate,
4010 `::help`. In this case, when the application starts, the runtime will
4011 simply output a list of loaded modules containing log expressions, then exit.
4013 #### Logging Expressions
4015 Rust provides several macros to log information. Here's a simple Rust program
4016 that demonstrates all four of them:
4020 #[phase(syntax, link)] extern crate log;
4023 error!("This is an error log")
4024 warn!("This is a warn log")
4025 info!("this is an info log")
4026 debug!("This is a debug log")
4030 These four log levels correspond to levels 1-4, as controlled by `RUST_LOG`:
4032 ``` {.bash .notrust}
4033 $ RUST_LOG=rust=3 ./rust
4034 This is an error log
4039 # Appendix: Rationales and design tradeoffs
4043 # Appendix: Influences and further references
4047 > The essential problem that must be solved in making a fault-tolerant
4048 > software system is therefore that of fault-isolation. Different programmers
4049 > will write different modules, some modules will be correct, others will have
4050 > errors. We do not want the errors in one module to adversely affect the
4051 > behaviour of a module which does not have any errors.
4053 > — Joe Armstrong
4055 > In our approach, all data is private to some process, and processes can
4056 > only communicate through communications channels. *Security*, as used
4057 > in this paper, is the property which guarantees that processes in a system
4058 > cannot affect each other except by explicit communication.
4060 > When security is absent, nothing which can be proven about a single module
4061 > in isolation can be guaranteed to hold when that module is embedded in a
4064 > — Robert Strom and Shaula Yemini
4066 > Concurrent and applicative programming complement each other. The
4067 > ability to send messages on channels provides I/O without side effects,
4068 > while the avoidance of shared data helps keep concurrent processes from
4073 Rust is not a particularly original language. It may however appear unusual
4074 by contemporary standards, as its design elements are drawn from a number of
4075 "historical" languages that have, with a few exceptions, fallen out of
4076 favour. Five prominent lineages contribute the most, though their influences
4077 have come and gone during the course of Rust's development:
4079 * The NIL (1981) and Hermes (1990) family. These languages were developed by
4080 Robert Strom, Shaula Yemini, David Bacon and others in their group at IBM
4081 Watson Research Center (Yorktown Heights, NY, USA).
4083 * The Erlang (1987) language, developed by Joe Armstrong, Robert Virding, Claes
4084 Wikström, Mike Williams and others in their group at the Ericsson Computer
4085 Science Laboratory (Älvsjö, Stockholm, Sweden) .
4087 * The Sather (1990) language, developed by Stephen Omohundro, Chu-Cheow Lim,
4088 Heinz Schmidt and others in their group at The International Computer
4089 Science Institute of the University of California, Berkeley (Berkeley, CA,
4092 * The Newsqueak (1988), Alef (1995), and Limbo (1996) family. These
4093 languages were developed by Rob Pike, Phil Winterbottom, Sean Dorward and
4094 others in their group at Bell Labs Computing Sciences Research Center
4095 (Murray Hill, NJ, USA).
4097 * The Napier (1985) and Napier88 (1988) family. These languages were
4098 developed by Malcolm Atkinson, Ron Morrison and others in their group at
4099 the University of St. Andrews (St. Andrews, Fife, UK).
4101 Additional specific influences can be seen from the following languages:
4103 * The structural algebraic types and compilation manager of SML.
4104 * The attribute and assembly systems of C#.
4105 * The references and deterministic destructor system of C++.
4106 * The memory region systems of the ML Kit and Cyclone.
4107 * The typeclass system of Haskell.
4108 * The lexical identifier rule of Python.
4109 * The block syntax of Ruby.
4111 [ffi]: guide-ffi.html