rust/doc/rust.md

% Rust Reference Manual

# Introduction

This document is the reference manual for the Rust programming language. It
provides three kinds of material:

  - Chapters that formally define the language grammar and, for each
    construct, informally describe its semantics and give examples of its
    use.
  - Chapters that informally describe the memory model, concurrency model,
    runtime services, linkage model and debugging facilities.
  - Appendix chapters providing rationale and references to languages that
    influenced the design.

This document does not serve as a tutorial introduction to the
language. Background familiarity with the language is assumed. A separate
tutorial document is available at <http://www.rust-lang.org/doc/tutorial>
to help acquire such background familiarity.

This document also does not serve as a reference to the core or standard
libraries included in the language distribution. Those libraries are
documented separately by extracting documentation attributes from their
source code. Formatted documentation can be found at the following
locations:

  - Core library: <http://doc.rust-lang.org/doc/core>
  - Standard library: <http://doc.rust-lang.org/doc/std>

## Disclaimer

Rust is a work in progress. The language continues to evolve as the design
shifts and is fleshed out in working code. Certain parts work, certain parts
do not, certain parts will be removed or changed.

This manual is a snapshot written in the present tense. All features described
exist in working code unless otherwise noted, but some are quite primitive or
remain to be further modified by planned work. Some may be temporary. It is a
*draft*, and we ask that you not take anything you read here as final.

If you have suggestions to make, please try to focus them on *reductions* to
the language: possible features that can be combined or omitted. We aim to
keep the size and complexity of the language under control.

# Notation

Rust's grammar is defined over Unicode codepoints, each conventionally
denoted `U+XXXX`, for 4 or more hexadecimal digits `X`. _Most_ of Rust's
grammar is confined to the ASCII range of Unicode, and is described in this
document by a dialect of Extended Backus-Naur Form (EBNF), specifically a
dialect of EBNF supported by common automated LL(k) parsing tools such as
`llgen`, rather than the dialect given in ISO 14977. The dialect can be
defined self-referentially as follows:

~~~~~~~~ {.ebnf .notation}

grammar : rule + ;
rule    : nonterminal ':' productionrule ';' ;
productionrule : production [ '|' production ] * ;
production : term * ;
term : element repeats ;
element : LITERAL | IDENTIFIER | '[' productionrule ']' ;
repeats : [ '*' | '+' ] NUMBER ? | NUMBER ? | '?' ;

~~~~~~~~

Where:

  - Whitespace in the grammar is ignored.
  - Square brackets are used to group rules.
  - `LITERAL` is a single printable ASCII character, or an escaped hexadecimal
     ASCII code of the form `\xQQ`, in single quotes, denoting the corresponding
     Unicode codepoint `U+00QQ`.
  - `IDENTIFIER` is a nonempty string of ASCII letters and underscores.
  - The `repeat` forms apply to the adjacent `element`, and are as follows:
    - `?` means zero or one repetition
    - `*` means zero or more repetitions
    - `+` means one or more repetitions
    - NUMBER trailing a repeat symbol gives a maximum repetition count
    - NUMBER on its own gives an exact repetition count

This EBNF dialect should hopefully be familiar to many readers.

The grammar for Rust given in this document is extracted and verified as LL(1)
by an automated grammar-analysis tool, and further tested against the Rust
sources. The generated parser is currently *not* the one used by the Rust
compiler itself, but in the future we hope to relate the two together more
precisely. As of this writing they are only related by testing against
existing source code.

## Unicode productions

A small number of productions in Rust's grammar permit Unicode codepoints
ouside the ASCII range; these productions are defined in terms of character
properties given by the Unicode standard, rather than ASCII-range
codepoints. These are given in the section [Special Unicode
Productions](#special-unicode-productions).

## String table productions

Some rules in the grammar -- notably [unary
operators](#unary-operator-expressions), [binary
operators](#binary-operator-expressions), [keywords](#keywords) and [reserved
words](#reserved-words) -- are given in a simplified form: as a listing of a
table of unquoted, printable whitespace-separated strings. These cases form a
subset of the rules regarding the [token](#tokens) rule, and are assumed to be
the result of a lexical-analysis phase feeding the parser, driven by a DFA,
operating over the disjunction of all such string table entries.

When such a string enclosed in double-quotes (`"`) occurs inside the
grammar, it is an implicit reference to a single member of such a string table
production. See [tokens](#tokens) for more information.


# Lexical structure

## Input format

Rust input is interpreted in as a sequence of Unicode codepoints encoded in
UTF-8. No normalization is performed during input processing. Most Rust
grammar rules are defined in terms of printable ASCII-range codepoints, but
a small number are defined in terms of Unicode properties or explicit
codepoint lists. ^[Surrogate definitions for the special Unicode productions
are provided to the grammar verifier, restricted to ASCII range, when
verifying the grammar in this document.]

## Special Unicode Productions

The following productions in the Rust grammar are defined in terms of
Unicode properties: `ident`, `non_null`, `non_star`, `non_eol`, `non_slash`,
`non_single_quote` and `non_double_quote`.

### Identifier

The `ident` production is any nonempty Unicode string of the following form:

   - The first character has property `XID_start`
   - The remaining characters have property `XID_continue`

that does _not_ occur in the set of [keywords](#keywords) or [reserved
words](#reserved-words).

Note: `XID_start` and `XID_continue` as character properties cover the
character ranges used to form the more familiar C and Java language-family
identifiers.

### Delimiter-restricted productions

Some productions are defined by exclusion of particular Unicode characters:

  - `non_null` is any single Unicode character aside from `U+0000` (null)
  - `non_eol` is `non_null` restricted to exclude `U+000A` (`'\n'`)
  - `non_star` is `non_null` restricted to exclude `U+002A` (`*`)
  - `non_slash` is `non_null` restricted to exclude `U+002F` (`/`)
  - `non_single_quote` is `non_null` restricted to exclude `U+0027`  (`'`)
  - `non_double_quote` is `non_null` restricted to exclude `U+0022` (`"`)

## Comments

~~~~~~~~ {.ebnf .gram}
comment : block_comment | line_comment ;
block_comment : "/*" block_comment_body * "*/" ;
block_comment_body : block_comment | non_star * | '*' non_slash ;
line_comment : "//" non_eol * ;
~~~~~~~~

Comments in Rust code follow the general C++ style of line and block-comment
forms, with proper nesting of block-comment delimiters. Comments are
interpreted as a form of whitespace.

## Whitespace

~~~~~~~~ {.ebnf .gram}
whitespace_char : '\x20' | '\x09' | '\x0a' | '\x0d' ;
whitespace : [ whitespace_char | comment ] + ;
~~~~~~~~

The `whitespace_char` production is any nonempty Unicode string consisting of any
of the following Unicode characters: `U+0020` (space, `' '`), `U+0009` (tab,
`'\t'`), `U+000A` (LF, `'\n'`), `U+000D` (CR, `'\r'`).

Rust is a "free-form" language, meaning that all forms of whitespace serve
only to separate _tokens_ in the grammar, and have no semantic meaning.

A Rust program has identical meaning if each whitespace element is replaced
with any other legal whitespace element, such as a single space character.

## Tokens

~~~~~~~~ {.ebnf .gram}
simple_token : keyword | reserved | unop | binop ; 
token : simple_token | ident | literal | symbol | whitespace token ;
~~~~~~~~

Tokens are primitive productions in the grammar defined by regular
(non-recursive) languages. "Simple" tokens are given in [string table
production](#string-table-productions) form, and occur in the rest of the
grammar as double-quoted strings. Other tokens have exact rules given.

### Keywords

The keywords in [crate files](#crate-files) are the following strings:

~~~~~~~~ {.keyword}
import export use mod dir
~~~~~~~~

The keywords in [source files](#source-files) are the following strings:

*TODO* split these between type keywords and regular (value) keywords,
 and define two different `identifier` productions for the different
 contexts.

~~~~~~~~ {.keyword}
alt any as assert
be bind block bool break
char check claim const cont
do
else export
f32 f64 fail false float fn for
i16 i32 i64 i8 if iface impl import in int
let log
mod mutable
native note
prove pure
resource ret
self str syntax
tag true type
u16 u32 u64 u8 uint unchecked unsafe use
vec
while
~~~~~~~~

Any of these have special meaning in their respective grammars, and are
excluded from the `ident` rule.

### Reserved words

The reserved words are the following strings:

~~~~~~~~ {.reserved}
m32 m64 m128
f80 f16 f128
class trait
~~~~~~~~

Any of these may have special meaning in future versions of the language, do
are excluded from the `ident` rule.

### Literals

A literal is an expression consisting of a single token, rather than a
sequence of tokens, that immediately and directly denotes the value it
evaluates to, rather than referring to it by name or some other evaluation
rule. A literal is a form of [constant expression](#constant-expression), so
is evaluated (primarily) at compile time.

~~~~~~~~ {.ebnf .gram}
literal : string_lit | char_lit | num_lit ;
~~~~~~~~

#### Character and string literals

~~~~~~~~ {.ebnf .gram}
char_lit : '\x27' char_body '\x27' ;
string_lit : '"' string_body * '"' ;

char_body : non_single_quote
          | '\x5c' [ '\x27' | common_escape ] ;

string_body : non_double_quote
            | '\x5c' [ '\x22' | common_escape ] ;

common_escape : '\x5c'
              | 'n' | 'r' | 't'
              | 'x' hex_digit 2
              | 'u' hex_digit 4
              | 'U' hex_digit 8 ;

hex_digit : 'a' | 'b' | 'c' | 'd' | 'e' | 'f'
          | 'A' | 'B' | 'C' | 'D' | 'E' | 'F'
          | dec_digit ;
dec_digit : '0' | nonzero_dec ;
nonzero_dec: '1' | '2' | '3' | '4'
           | '5' | '6' | '7' | '8' | '9' ;
~~~~~~~~

A _character literal_ is a single Unicode character enclosed within two
`U+0027` (single-quote) characters, with the exception of `U+0027` itself,
which must be _escaped_ by a preceding U+005C character (`\`).

A _string literal_ is a sequence of any Unicode characters enclosed within
two `U+0022` (double-quote) characters, with the exception of `U+0022`
itself, which must be _escaped_ by a preceding `U+005C` character (`\`).

Some additional _escapes_ are available in either character or string
literals. An escape starts with a `U+005C` (`\`) and continues with one of
the following forms:

  * An _8-bit codepoint escape_ escape starts with `U+0078` (`x`) and is
    followed by exactly two _hex digits_. It denotes the Unicode codepoint
    equal to the provided hex value.
  * A _16-bit codepoint escape_ starts with `U+0075` (`u`) and is followed
    by exactly four _hex digits_. It denotes the Unicode codepoint equal to
    the provided hex value.
  * A _32-bit codepoint escape_ starts with `U+0055` (`U`) and is followed
    by exactly eight _hex digits_. It denotes the Unicode codepoint equal to
    the provided hex value.
  * A _whitespace escape_ is one of the characters `U+006E` (`n`), `U+0072`
    (`r`), or `U+0074` (`t`), denoting the unicode values `U+000A` (LF),
    `U+000D` (CR) or `U+0009` (HT) respectively.
  * The _backslash escape_ is the character U+005C (`\`) which must be
    escaped in order to denote *itself*.

#### Number literals

~~~~~~~~ {.ebnf .gram}

num_lit : nonzero_dec [ dec_digit | '_' ] * num_suffix ?
        | '0' [       [ dec_digit | '_' ] + num_suffix ?
              | 'b'   [ '1' | '0' | '_' ] + int_suffix ?
              | 'x'   [ hex_digit | '-' ] + int_suffix ? ] ;

num_suffix : int_suffix | float_suffix ;

int_suffix : 'u' int_suffix_size ?
           | 'i' int_suffix_size ;
int_suffix_size : [ '8' | '1' '6' | '3' '2' | '6' '4' ] ;

float_suffix : [ exponent | '.' dec_lit exponent ? ] float_suffix_ty ? ;
float_suffix_ty : 'f' [ '3' '2' | '6' '4' ] ;
exponent : ['E' | 'e'] ['-' | '+' ] ? dec_lit ;
dec_lit : [ dec_digit | '_' ] + ;
~~~~~~~~

A _number literal_ is either an _integer literal_ or a _floating-point
literal_. The grammar for recognizing the two kinds of literals is mixed
as they are differentiated by suffixes.

##### Integer literals

An _integer literal_ has one of three forms:

  * A _decimal literal_ starts with a *decimal digit* and continues with any
    mixture of *decimal digits* and _underscores_.
  * A _hex literal_ starts with the character sequence `U+0030` `U+0078`
    (`0x`) and continues as any mixture hex digits and underscores.
  * A _binary literal_ starts with the character sequence `U+0030` `U+0062`
    (`0b`) and continues as any mixture binary digits and underscores.

By default, an integer literal is of type `int`. An integer literal may be
followed (immediately, without any spaces) by an _integer suffix_, which
changes the type of the literal. There are two kinds of integer literal
suffix:

  * The `u` suffix gives the literal type `uint`.
  * Each of the signed and unsigned machine types `u8`, `i8`,
    `u16`, `i16`, `u32`, `i32`, `u64` and `i64`
    give the literal the corresponding machine type.


Examples of integer literals of various forms:

~~~~
123;                               // type int
123u;                              // type uint
123_u;                             // type uint
0xff00;                            // type int
0xff_u8;                           // type u8
0b1111_1111_1001_0000_i32;         // type i32
~~~~

##### Floating-point literals

A _floating-point literal_ has one of two forms:

* Two _decimal literals_ separated by a period
  character `U+002E` (`.`), with an optional _exponent_ trailing after the
  second decimal literal.
* A single _decimal literal_ followed by an _exponent_.

By default, a floating-point literal is of type `float`. A
floating-point literal may be followed (immediately, without any
spaces) by a _floating-point suffix_, which changes the type of the
literal. There are three floating-point suffixes: `f` (for the base
`float` type), `f32`, and `f64` (the 32-bit and 64-bit floating point
types).

A set of suffixes are also reserved to accommodate literal support for
types corresponding to reserved tokens. The reserved suffixes are `f16`,
`f80`, `f128`, `m`, `m32`, `m64` and `m128`.

Examples of floating-point literals of various forms:

~~~~
123.0;                             // type float
0.1;                               // type float
3f;                                // type float
0.1f32;                            // type f32
12E+99_f64;                        // type f64
~~~~

##### Nil and boolean literals

The _nil value_, the only value of the type by the same name, is
written as `()`. The two values of the boolean type are written `true`
and `false`.

### Symbols

~~~~~~~~ {.ebnf .gram}
symbol : "::" "->"
       | '#' | '[' | ']' | '(' | ')' | '{' | '}'
       | ',' | ';' ;
~~~~~~~~

Symbols are a general class of printable [token](#tokens) that play structural
roles in a variety of grammar productions. They are catalogued here for
completeness as the set of remaining miscellaneous printable token that do not
otherwise appear as [unary operators](#unary-operator-expressions), [binary
operators](#binary-operator-expressions), [keywords](#keywords) or [reserved
words](#reserved-words).


## Paths

~~~~~~~~ {.ebnf .gram}

expr_path : ident [ "::" expr_path_tail ] + ;
expr_path_tail : '<' type_expr [ ',' type_expr ] + '>'
               | expr_path ;

type_path : ident [ type_path_tail ] + ;
type_path_tail : '<' type_expr [ ',' type_expr ] + '>'
               | "::" type_path ;

~~~~~~~~

A _path_ is a sequence of one or more path components _logically_ separated by
a namespace qualifier (`::`). If a path consists of only one component, it
may refer to either an [item](#items) or a (variable)[#variables) in a local
control scope. If a path has multiple components, it refers to an item.

Every item has a _canonical path_ within its [crate](#crates), but the path
naming an item is only meaningful within a given crate. There is no global
namespace across crates; an item's canonical path merely identifies it within
the crate.

Two examples of simple paths consisting of only identifier components:

~~~~
x;
x::y::z;
~~~~

Path components are usually [identifiers](#identifiers), but the trailing
component of a path may be an angle-bracket enclosed list of [type
arguments](type-arguments). In [expression](#expressions) context, the type
argument list is given after a final (`::`) namespace qualifier in order to
disambiguate it from a relational expression involving the less-than symbol
(`<`). In [type expression](#type-expressions) context, the final namespace
qualifier is omitted.

Two examples of paths with type arguments:

~~~~
type t = map::hashtbl<int,str>;  // Type arguments used in a type expression
let x = id::<int>(10);           // Type arguments used in a call expression
~~~~


# Crates and source files

Rust is a *compiled* language. Its semantics are divided along a
*phase distinction* between compile-time and run-time. Those semantic
rules that have a *static interpretation* govern the success or failure
of compilation. A program that fails to compile due to violation of a
compile-time rule has no defined semantics at run-time; the compiler should
halt with an error report, and produce no executable artifact.

The compilation model centres on artifacts called _crates_. Each compilation
is directed towards 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 a unit of compilation and linking, as well as versioning,
distribution and runtime loading. A crate contains a _tree_ of nested
[module](#modules) scopes. 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.

Crates are provided to the Rust compiler through two kinds of file:

  - _crate files_, that end in `.rc` and each define a `crate`.
  - _source files_, that end in `.rs` and each define a `module`.

The Rust compiler is always invoked with a single input file, and always
produces a single output crate.

When the Rust compiler is invoked with a crate file, it reads the _explicit_
definition of the crate it's compiling from that file, and populates the
crate with modules derived from all the source files referenced by the
crate, reading and processing all the referenced modules at once.

When the Rust compiler is invoked with a source file, it creates an
_implicit_ crate and treats the source file and though it was referenced as
the sole module populating this implicit crate. The module name is derived
from the source file name, with the `.rs` extension removed.

## Crate files

~~~~~~~~ {.ebnf .gram}
crate : attribute [ ';' | attribute* directive ] 
      | directive ;
directive : view_item | dir_directive | source_directive ;
~~~~~~~~

A crate file contains a crate definition, for which the production above
defines the grammar. It is a declarative grammar that guides the compiler in
assembling a crate from component source files.^[A crate is somewhat
analogous to an *assembly* in the ECMA-335 CLI model, a *library* in the
SML/NJ Compilation Manager, a *unit* in the Owens and Flatt module system,
or a *configuration* in Mesa.] A crate file describes:

* [Attributes](#attributes) about the crate, such as author, name, version,
  and copyright. These are used for linking, versioning and distributing
  crates.
* The source-file and directory modules that make up the crate.
* Any `use`, `import` or `export` [view items](#view-items) that apply to the
  anonymous module at the top-level of the crate's module tree.

An example of a crate file:

~~~~~~~~
// Linkage attributes
#[ link(name = "projx"
        vers = "2.5",
        uuid = "9cccc5d5-aceb-4af5-8285-811211826b82") ];

// Additional metadata attributes
#[ desc = "Project X",
   license = "BSD" ];
   author = "Jane Doe" ];

// Import a module.
use std (ver = "1.0");

// Define some modules.
#[path = "foo.rs"]
mod foo;
mod bar {
    #[path =  "quux.rs"]
    mod quux;
}
~~~~~~~~

### Dir directives

A `dir_directive` forms a module in the module tree making up the crate, as
well as implicitly relating that module to a directory in the filesystem
containing source files and/or further subdirectories. The filesystem
directory associated with a `dir_directive` module can either be explicit,
or if omitted, is implicitly the same name as the module.

A `source_directive` references a source file, either explicitly or
implicitly by combining the module name with the file extension `.rs`.  The
module contained in that source file is bound to the module path formed by
the `dir_directive` modules containing the `source_directive`.

## Source file

A source file contains a `module`, that is, a sequence of zero-or-more
`item` definitions. Each source file is an implicit module, the name and
location of which -- in the module tree of the current crate -- is defined
from outside the source file: either by an explicit `source_directive` in
a referencing crate file, or by the filename of the source file itself.


# Items and attributes

A crate is a collection of [items](#items), each of which may have some number
of [attributes](#attributes) attached to it.

## Items

~~~~~~~~ {.ebnf .gram}
item : mod_item | fn_item | type_item | enum_item
     | res_item | iface_item | impl_item ;
~~~~~~~~

An _item_ is a component of a crate; some module items can be defined in
[crate files](#crate-files) but most are defined in [source
files](#source-files). Items are organized within a crate by a nested set of
[modules](#modules). Every crate has a single "outermost" anonymous module;
all further items within the crate have [paths](#paths) within the module tree
of the crate.

Items are entirely determined at compile-time, remain constant during
execution, and may reside in read-only memory.

There are several kinds of item:

  * [modules](#modules)
  * [functions](#functions)
  * [type definitions](#type-definitions)
  * [eunmerations](#enumerations)
  * [resources](#resources)
  * [interfaces](#interfaces)
  * [implementations](#implementations)

Some items form an implicit scope for the declaration of sub-items. In other
words, within a function or module, declarations of items can (in many cases)
be mixed with the statements, control blocks, and similar artifacts that
otherwise compose the item body. The meaning of these scoped items is the same
as if the item was declared outside the scope -- it is still a static item --
except that the item's *path name* within the module namespace is qualified by
the name of the enclosing item, or is private to the enclosing item (in the
case of functions). The exact locations in which sub-items may be declared is
given by the grammar.

All items except modules may be *parametrized* by type. Type parameters are
given as a comma-separated list of identifiers enclosed in angle brackets
(`<...>`), after the name of the item and before its definition. The type
parameters of an item are considered "part of the name", not the type of the
item; in order to refer to the type-parametrized item, a referencing
[path](#paths) must in general provide type arguments as a list of
comma-separated types enclosed within angle brackets. In practice, the
type-inference system can usually infer such argument types from
context. There are no general type-parametric types, only type-parametric
items.


### Modules

~~~~~~~~ {.ebnf .gram}
mod_item : "mod" ident '{' mod '}' ;
mod : [ view_item | item ] * ;
~~~~~~~~

A module is a container for zero or more [view items](#view-items) and zero or
more [items](#items). The view items manage the visibility of the items
defined within the module, as well as the visibility of names from outside the
module when referenced from inside the module.

A _module item_ is a module, surrounded in braces, named, and prefixed with
the keyword `mod`. A module item introduces a new, named module into the tree
of modules making up a crate. Modules can nest arbitrarily.

An example of a module:

~~~~~~~~
mod math {
    type complex = (f64,f64);
    fn sin(f64) -> f64 {
        ...
    }
    fn cos(f64) -> f64 {
        ...
    }
    fn tan(f64) -> f64 {
        ...
    }
    ...
}
~~~~~~~~


#### View items

~~~~~~~~ {.ebnf .gram}
view_item : use_decl | import_decl | export_decl ;
~~~~~~~~

A view item manages the namespace of a module; it does not define new items
but simply changes the visibility of other items. There are several kinds of
view item:

 * [use declarations](#use-declarations)
 * [import declarations](#import-declarations)
 * [export declarations](#export-declarations)

##### Use declarations

~~~~~~~~ {.ebnf .gram}
use_decl : "use" ident [ '(' link_attrs ')' ] ? ;
link_attrs : link_attr [ ',' link_attrs ] + ;
link_attr : ident '=' literal ;
~~~~~~~~

A _use declaration_ specifies a dependency on an external crate. The external
crate is then imported into the declaring scope as the `ident` provided in the
`use_decl`.

The external crate is resolved to a specific `soname` at compile time, and a
runtime linkage requirement to that `soname` is passed to the linker for
loading at runtime. The `soname` is resolved at compile time by scanning the
compiler's library path and matching the `link_attrs` provided in the
`use_decl` against any `#link` attributes that were declared on the external
crate when it was compiled. If no `link_attrs` are provided, a default `name`
attribute is assumed, equal to the `ident` given in the `use_decl`.

Two examples of `use` declarations:

~~~~~~~~
use pcre (uuid = "54aba0f8-a7b1-4beb-92f1-4cf625264841");

use std; // equivalent to: use std ( name = "std" );

use ruststd (name = "std"); // linking to 'std' under another name
~~~~~~~~

##### Import declarations

~~~~~~~~ {.ebnf .gram}
import_decl : "import" ident [ '=' path
                             | "::" path_glob ] ;

path_glob : ident [ "::" path_glob ] ?
          | '*'
          | '{' ident [ ',' ident ] * '}'
~~~~~~~~

An _import declaration_ creates one or more local name bindings synonymous
with some other [path](#paths). Usually an import declaration is used to
shorten the path required to refer to a module item.

*Note*: unlike many languages, Rust's `import` declarations do not* declare
*linkage-dependency with external crates. Linkage dependencies are
*independently declared with [`use` declarations](#use -eclarations).

Imports support a number of "convenience" notations:

  * Importing as a different name than the imported name, using the 
    syntax `import x = p::q::r;`.
  * Importing a list of paths differing only in final element, using
    the glob-like brace syntax `import a::b::{c,d,e,f};`
  * Importing all paths matching a given prefix, using the glob-like
    asterisk syntax `import a::b::*;`

An example of imports:

~~~~
import foo = core::info;
import std::math::sin;
import std::str::{char_at, hash};
import core::option::*;

fn main() {
    // Equivalent to 'log(core::info, std::math::sin(1.0));'
    log(foo, sin(1.0));

    // Equivalent to 'log(core::info, core::option::some(1.0));'
    log(info, some(1.0));

    // Equivalent to 'log(core::info,
    //                    std::str::hash(std::str::char_at("foo")));'
    log(info, hash(char_at("foo")));
}
~~~~

##### Export declarations

~~~~~~~~ {.ebnf .gram}
export_decl : "export" ident [ ',' ident ] * ;
~~~~~~~~

An _export declaration_ restricts the set of local names within a module that
can be accessed from code outside the module. By default, all _local items_ in
a module are exported; imported paths are not automatically re-exported by
default. If a module contains an explicit `export` declaration, this
declaration replaces the default export with the export specified.

An example of an export:

~~~~~~~~
mod foo {
    export primary;

    fn primary() {
        helper(1, 2);
        helper(3, 4);
    }

    fn helper(x: int, y: int) {
        ...
    }
}

fn main() {
    foo::primary();  // Will compile.
    foo::helper(2,3) // ERROR: will not compile.
}
~~~~~~~~

Multiple names may be exported from a single export declaration:

~~~~~~~~
mod foo {
    export primary, secondary;

    fn primary() {
        helper(1, 2);
        helper(3, 4);
    }

    fn secondary() {
        ...
    }

    fn helper(x: int, y: int) {
        ...
    }
}
~~~~~~~~


### Functions

A _function item_ defines a sequence of [statements](#statements) and an
optional final [expression](#expressions) associated with a name and a set of
parameters. Functions are declared with the keyword `fn`. Functions declare a
set of *input [slots](#slots)* as parameters, through which the caller passes
arguments into the function, and an *output [slot](#slots)* through which the
function passes results back to the caller.

A function may also be copied into a first class *value*, in which case the
value has the corresponding [*function type*](#function-types), and can be
used otherwise exactly as a function item (with a minor additional cost of
calling the function, as such a call is indirect).

Every control path in a function logically ends with a `ret` expression or a
diverging expression. If the outermost block of a function has a
value-producing expression in its final-expression position, that expression
is interpreted as an implicit `ret` expression applied to the
final-expression.

An example of a function:

~~~~
fn add(x: int, y: int) -> int {
    ret x + y;
}
~~~~

#### Diverging functions

A special kind of function can be declared with a `!` character where the
output slot type would normally be. For example:

~~~~
fn my_err(s: str) -> ! {
    log(info, s);
    fail;
}
~~~~

We call such functions "diverging" because they never return a value to the
caller. Every control path in a diverging function must end with a
[`fail`](#fail-expressions) or a call to another diverging function on every
control path. The `!` annotation does *not* denote a type. Rather, the result
type of a diverging function is a special type called $\bot$ ("bottom") that
unifies with any type. Rust has no syntax for $\bot$.

It might be necessary to declare a diverging function because as mentioned
previously, the typechecker checks that every control path in a function ends
with a [`ret`](#return-expressions) or diverging expression. So, if `my_err`
were declared without the `!` annotation, the following code would not
typecheck:

~~~~
fn f(i: int) -> int {
   if i == 42 {
     ret 42;
   }
   else {
     my_err("Bad number!");
   }
}
~~~~

The typechecker would complain that `f` doesn't return a value in the
`else` branch. Adding the `!` annotation on `my_err` would
express that `f` requires no explicit `ret`, as if it returns
control to the caller, it returns a value (true because it never returns
control).

#### Predicate functions

Any pure boolean function is called a *predicate function*, and may be used in
a [constraint](#constraints), as part of the static [typestate
system](#typestate-system). A predicate declaration is identical to a function
declaration, except that it is declared with the additional keyword `pure`. In
addition, the typechecker checks the body of a predicate with a restricted set
of typechecking rules. A predicate

* may not contain an assignment or self-call expression; and
* may only call other predicates, not general functions.

An example of a predicate:

~~~~
pure fn lt_42(x: int) -> bool {
    ret (x < 42);
}
~~~~

A non-boolean function may also be declared with `pure fn`. This allows
predicates to call non-boolean functions as long as they are pure. For example:

~~~~
pure fn pure_length<T>(ls: list<T>) -> uint { /* ... */ }

pure fn nonempty_list<T>(ls: list<T>) -> bool { pure_length(ls) > 0u }
~~~~

In this example, `nonempty_list` is a predicate---it can be used in a
typestate constraint---but the auxiliary function `pure_length`@ is
not.

*ToDo:* should actually define referential transparency.

The effect checking rules previously enumerated are a restricted set of
typechecking rules meant to approximate the universe of observably
referentially transparent Rust procedures conservatively. Sometimes, these
rules are *too* restrictive. Rust allows programmers to violate these rules by
writing predicates that the compiler cannot prove to be referentially
transparent, using an escape-hatch feature called "unchecked blocks". When
writing code that uses unchecked blocks, programmers should always be aware
that they have an obligation to show that the code *behaves* referentially
transparently at all times, even if the compiler cannot *prove* automatically
that the code is referentially transparent. In the presence of unchecked
blocks, the compiler provides no static guarantee that the code will behave as
expected at runtime. Rather, the programmer has an independent obligation to
verify the semantics of the predicates they write.

*ToDo:* last two sentences are vague.

An example of a predicate that uses an unchecked block:

~~~~
fn pure_foldl<T, U: copy>(ls: list<T>, u: U, f: block(&T, &U) -> U) -> U {
    alt ls {
      nil. { u }
      cons(hd, tl) { f(hd, pure_foldl(*tl, f(hd, u), f)) }
    }
}

pure fn pure_length<T>(ls: list<T>) -> uint {
    fn count<T>(_t: T, u: uint) -> uint { u + 1u }
    unchecked {
        pure_foldl(ls, 0u, count)
    }
}
~~~~

Despite its name, `pure_foldl` is a `fn`, not a `pure fn`, because there is no
way in Rust to specify that the higher-order function argument `f` is a pure
function. So, to use `foldl` in a pure list length function that a predicate
could then use, we must use an `unchecked` block wrapped around the call to
`pure_foldl` in the definition of `pure_length`.

#### Generic functions

A _generic function_ allows one or more _parameterized types_ to
appear in its signature. Each type parameter must be explicitly
declared, in an angle-bracket-enclosed, comma-separated list following
the function name.

~~~~
fn iter<T>(seq: [T], f: block(T)) {
    for elt: T in seq { f(elt); }
}
fn map<T, U>(seq: [T], f: block(T) -> U) -> [U] {
    let acc = [];
    for elt in seq { acc += [f(elt)]; }
    acc
}
~~~~

Inside the function signature and body, the name of the type parameter
can be used as a type name.

When a generic function is referenced, its type is instantiated based
on the context of the reference. For example, calling the `iter`
function defined above on `[1, 2]` will instantiate type parameter `T`
with `int`, and require the closure parameter to have type
`block(int)`.

Since a parameter type is opaque to the generic function, the set of
operations that can be performed on it is limited. Values of parameter
type can always be moved, but they can only be copied when the
parameter is given a [`copy` bound](#type-kinds).

~~~~
fn id<T: copy>(x: T) -> T { x }
~~~~

Similarly, [interface](#interfaces) bounds can be specified for type
parameters to allow methods of that interface to be called on values
of that type.

### Type definitions

A _type definition_ defines a new name for an existing [type](#types). Type
definitions are declared with the keyword `type`. Every value has a single,
specific type; the type-specified aspects of a value include:

* Whether the value is composed of sub-values or is indivisible.
* Whether the value represents textual or numerical information.
* Whether the value represents integral or floating-point information.
* The sequence of memory operations required to access the value.
* The *kind* of the type (pinned, unique or shared).

For example, the type `{x: u8, y: u8@`} defines the set of immutable values
that are composite records, each containing two unsigned 8-bit integers
accessed through the components `x` and `y`, and laid out in memory with the
`x` component preceding the `y` component.

### Enumerations

An _enumeration item_ simultaneously declares a new nominal [enumerated
type](#enumerated-type) as well as a set of *constructors* that can be used to
create or pattern-match values of the corresponding enumerated type.

The constructors of an `enum` type may be recursive: that is, each constructor
may take an argument that refers, directly or indirectly, to the enumerated
type the constructor is a member of. Such recursion has restrictions:

* Recursive types can be introduced only through `enum` constructors.
* A recursive `enum` item must have at least one non-recursive constructor (in
  order to give the recursion a basis case).
* The recursive argument of recursive `enum` constructors must be [*box*
  values](#box-types) (in order to bound the in-memory size of the
  constructor).
* Recursive type definitions can cross module boundaries, but not module
  *visibility* boundaries, nor crate boundaries (in order to simplify the
  module system).


An example of an `enum` item and its use:

~~~~
enum animal {
  dog;
  cat;
}

let a: animal = dog;
a = cat;
~~~~

An example of a *recursive* `enum` item and its use:

~~~~
enum list<T> {
  nil;
  cons(T, @list<T>);
}

let a: list<int> = cons(7, @cons(13, @nil));
~~~~

### Resources

_Resources_ are values that have a destructor associated with them. A
_resource item_ is used to declare resource type and constructor.

~~~~
resource file_descriptor(fd: int) {
    std::os::libc::close(fd);
}
~~~~

Calling the `file_descriptor` constructor function on an integer will
produce a value with the `file_descriptor` type. Resource types have a
noncopyable [type kind](#type-kinds), and thus may not be copied. It
is guaranteed, barring drastic program termination that somehow
prevents unwinding from taking place, that for each constucted
resource value, the destructor will run once, when the value is
disposed of. For stack-allocated values, this happens when the value
goes out of scope. For values in shared boxes, it happens when the
reference count of the box reaches zero.

The argument to the resource constructor is stored in the resulting
value, and can be accessed using the dereference (`*`) [unary
operator](#unary-operator-expressions).

### Interfaces

An _interface item_ describes a set of method types. _[implementation
items](#implementations)_ can be used to provide implementations of
those methods for a specific type.

~~~~
iface shape {
    fn draw(surface);
    fn bounding_box() -> bounding_box;
}
~~~~

This defines an interface with two methods. All values which have
[implementations](#implementations) of this interface in scope can
have their `draw` and `bounding_box` methods called, using
`value.bounding_box()` [syntax](#field-expressions).

Type parameters can be specified for an interface to make it generic.
These appear after the name, using the same syntax used in [generic
functions](#generic-functions).

~~~~
iface seq<T> {
   fn len() -> uint;
   fn elt_at(n: uint) -> T;
   fn iter(block(T));
}
~~~~

Generic functions may use interfaces as bounds on their type
parameters. This will have two effects: only types that implement the
interface can be used to instantiate the parameter, and within the
generic function, the methods of the interface can be called on values
that have the parameter's type. For example:

~~~~
fn draw_twice<T: shape>(surface: surface, sh: T) {
    sh.draw(surface);
    sh.draw(surface);
}
~~~~

Interface items also define a type with the same name as the
interface. Values of this type are created by
[casting](#type-cast-expressions) values (of a type for which an
implementation of the given interface is in scope) to the interface
type.

~~~~
let myshape: shape = mycircle as shape;
~~~~

The resulting value is a reference counted box containing the value
that was cast along with information that identify the methods of the
implementation that was used. Values with an interface type can always
have methods of their interface called on them, and can be used to
instantiate type parameters that are bounded on their interface.

### Implementations

An _implementation item_ provides an implementation of an
[interfaces](#interfaces) for a type.

~~~~
type circle = {radius: float, center: point};
    
impl circle_shape of shape for circle {
    fn draw(s: surface) { do_draw_circle(s, self); }
    fn bounding_box() -> bounding_box {
        let r = self.radius;
        {x: self.center.x - r, y: self.center.y - r,
         width: 2 * r, height: 2 * r}
    }
}
~~~~

This defines an implementation named `circle_shape` of interface
`shape` for type `circle`. The name of the implementation is the name
by which it is imported and exported, but has no further significance.
It may be left off to default to the name of the interface that was
implemented. Implementation names do not conflict the way other names
do—multiple implementations with the same name may exist in a scope at
the same time.

It is possible to define an implementation without referencing an
interface. The methods in such an implementation can only be used
statically (as direct calls on the values of the type that the
implementation targets). In such an implementation, the `of` clause is
not given, and the name is mandatory.

~~~~
impl uint_loops for uint {
    fn times(f: block(uint)) {
        let i = 0;
        while i < self { f(i); i += 1u; }
    }
}
~~~~

_When_ an interface is specified, all methods declared as part of the
interface must be present, with matching types and type parameter
counts, in the implementation.

An implementation can take type parameters, which can be different
from the type parameters taken by the interface it implements. They
are written after the name of the implementation, or if that is not
specified, after the `impl` keyword.

~~~~
impl <T> of seq<T> for [T] {
    /* ... */
}
impl of seq<bool> for u32 {
   /* Treat the integer as a sequence of bits */
}
~~~~

## Attributes

~~~~~~~~{.ebnf .gram}
attribute : '#' '[' attr_list ']' ;
attr_list : attr [ ',' attr_list ]*
attr : ident [ '=' literal
             | '(' attr_list ')' ] ? ;
~~~~~~~~

Static entities in Rust -- crates, modules and items -- may have _attributes_
applied to them. ^[Attributes in Rust are modeled on Attributes in ECMA-335,
C#] An attribute is a general, free-form piece of metadata that is interpreted
according to name, convention, and language and compiler version.  Attributes
may appear as any of:

* A single identifier, the attribute name
* An identifier followed by the equals sign '=' and a literal, providing a key/value pair
* An identifier followed by a parenthesized list of sub-attribute arguments

Attributes are applied to an entity by placing them within a hash-list
(`#[...]`) as either a prefix to the entity or as a semicolon-delimited
declaration within the entity body.

An example of attributes:

~~~~~~~~
// A function marked as a unit test
#[test]
fn test_foo() {
  ...
}

// General metadata applied to the enclosing module or crate.
#[license = "BSD"];

// A conditionally-compiled module
#[cfg(target_os="linux")]
mod bar {
  ...
}

// A documentation attribute
#[doc = "Add two numbers together."
fn add(x: int, y: int) { x + y }
~~~~~~~~

In future versions of Rust, user-provided extensions to the compiler will be
able to interpret attributes. When this facility is provided, a distinction
will be made between language-reserved and user-available attributes.

At present, only the Rust compiler interprets attributes, so all attribute
names are effectively reserved. Some significant attributes include:

* The `doc` attribute, for documenting code where it's written.
* The `cfg` attribute, for conditional-compilation by build-configuration.
* The `link` attribute, for describing linkage metadata for a crate.
* The `test` attribute, for marking functions as unit tests.

Other attributes may be added or removed during development of the language.


# Statements and expressions

Rust is _primarily_ an expression language. This means that most forms of
value-producing or effect-causing evaluation are directed by the uniform
syntax category of _expressions_. Each kind of expression can typically _nest_
within each other kind of expression, and rules for evaluation of expressions
involve specifying both the value produced by the expression and the order in
which its sub-expressions are themselves evaluated.

In contrast, statements in Rust serve _mostly_ to contain and explicitly
sequence expression evaluation.

## Statements

A _statement_ is a component of a block, which is in turn a component of an
outer [block-expression](#block-expressions) or [function](#functions). When a
function is spawned into a [task](#tasks), the task *executes* statements in
an order determined by the body of the enclosing function. Each statement
causes the task to perform certain actions.

Rust has two kinds of statement: [declarations](#declarations) and
[expressions](#expressions).

A declaration serves to introduce a *name* that can be used in the block
*scope* enclosing the statement: all statements before and after the
name, from the previous opening curly-brace (`{`) up to the next closing
curly-brace (`}`).

An expression serves the dual roles of causing side effects and producing a
*value*. Expressions are said to *evaluate to* a value, and the side effects
are caused during *evaluation*. Many expressions contain sub-expressions as
operands; the definition of each kind of expression dictates whether or not,
and in which order, it will evaluate its sub-expressions, and how the
expression's value derives from the value of its sub-expressions.

In this way, the structure of execution -- both the overall sequence of
observable side effects and the final produced value -- is dictated by the
structure of expressions. Blocks themselves are expressions, so the nesting
sequence of block, statement, expression, and block can repeatedly nest to an
arbitrary depth.

### Declaration statements

A _declaration statement_ is one that introduces a *name* into the
enclosing statement block. The declared name may denote a new slot or a new
item.

#### Item declarations

An _item declaration statement_ has a syntactic form identical to an
[item](#items) declaration within a module. Declaring an item -- a function,
enumeration, type, resource, interface, implementation or module -- locally
within a statement block is simply a way of restricting its scope to a narrow
region containing all of its uses; it is otherwise identical in meaning to
declaring the item outside the statement block.

Note: there is no implicit capture of the function's dynamic environment when
declaring a function-local item.


#### Slot declarations

A _slot declaration_ has one one of two forms:

* `let` `pattern` `optional-init`;
* `let` `pattern` : `type` `optional-init`;

Where `type` is a type expression, `pattern` is an irrefutable pattern (often
just the name of a single slot), and `optional-init` is an optional
initializer. If present, the initializer consists of either an assignment
operator (`=`) or move operator (`<-`), followed by an expression.

Both forms introduce a new slot into the enclosing block scope. The new slot
is visible from the point of declaration until the end of the enclosing block
scope.

The former form, with no type annotation, causes the compiler to infer the
static type of the slot through unification with the types of values assigned
to the slot in the remaining code in the block scope. Inference only occurs on
frame-local slots, not argument slots. Function signatures must
always declared types for all argument slots.


### Expression statements

An _expression statement_ is one that evaluates an [expression](#expressions)
and drops its result. The purpose of an expression statement is often to cause
the side effects of the expression's evaluation.

## Expressions

### Literal expressions

A _literal expression_ consists of one of the [literal](#literals)
forms described earlier. It directly describes a number, character,
string, boolean value, or the nil value.

~~~~~~~~ {.literals}
();        // nil type
"hello";   // string type
'5';       // character type
5;         // integer type
~~~~~~~~

### Tuple expressions

Tuples are written by enclosing two or more comma-separated
expressions in parentheses. They are used to create [tuple
typed](#tuple-types) values.

~~~~~~~~ {.tuple}
(0f, 4.5f);
("a", 4u, true)
~~~~~~~~

### Record expressions

A _[record](#record-types) expression_ is one or more comma-separated
name-value pairs enclosed by braces. A fieldname can be any identifier
(including reserved words), and is separated from its value expression
by a colon. To indicate that a field is mutable, the `mutable` keyword
is written before its name.

~~~~
{x: 10f, y: 20f};
{name: "Joe", age: 35u, score: 100_000};
{ident: "X", mutable count: 0u};
~~~~

The order of the fields in a record expression is significant, and
determines the type of the resulting value. `{a: u8, b: u8}` and `{b:
u8, a: u8}` are two different fields.

A record expression can be ended with the word `with` followed by an
expression to denote a functional update. The expression following
`with` (the base) be of a record type that includes at least all the
fields mentioned in the record expression. A new record will be
created, of the same type as the base expression, with the given
values for the fields that were explicitly specified, and the values
in the base record for all other fields. The ordering of the fields in
such a record expression is not significant.

~~~~
let base = {x: 1, y: 2, z: 3};
{y: 0, z: 10 with base};
~~~~

### Field expressions

A dot can be used to access a field in a record.

~~~~~~~~ {.field}
myrecord.myfield;
{a: 10, b: 20}.a;
~~~~~~~~

A field access on a record is an _lval_ referring to the value of that
field. When the field is mutable, it can be
[assigned](#assignment-expressions) to.

When the type of the expression to the left of the dot is a boxed
record, it is automatically derferenced to make the field access
possible.

Field access syntax is overloaded for [interface method](#interfaces)
access. When no matching field is found, or the expression to the left
of the dot is not a (boxed) record, an
[implementation](#implementations) that matches this type and the
given method name is looked up instead, and the result of the
expression is this method, with its _self_ argument bound to the
expression on the left of the dot.

### Vector expressions

A _[vector](#vector-types) expression_ is written by enclosing zero or
more comma-separated expressions of uniform type in square brackets.
The keyword `mutable` can be written after the opening bracket to
indicate that the elements of the resulting vector may be mutated.
When no mutability is specified, the vector is immutable.

~~~~
[1, 2, 3, 4];
["a", "b", "c", "d"];
[mutable 0u8, 0u8, 0u8, 0u8];
~~~~

### Index expressions

[Vector](#vector-types)-typed expressions can be indexed by writing a
square-bracket-enclosed expression (the index) after them. When the
vector is mutable, the resulting _lval_ can be assigned to.

Indices are zero-based, and may be of any integral type. Vector access
is bounds-checked at run-time. When the check fails, it will put the
task in a _failing state_.

~~~~
[1, 2, 3, 4][0];
[mutable 'x', 'y'][1] = 'z';
["a", "b"][10]; // fails
~~~~

### Unary operator expressions

Rust defines five unary operators. They are all written a prefix
operators, before the expression they apply to.

`-`
  : Negation. May only be applied to numeric types.
`*`
  : Dereference. When applied to a [box](#box-types) or
    [resource](#resources) type, it accesses the inner value. For
    mutable boxes, the resulting _lval_ can be assigned to. For
    [enums](#enumerated-types) that have only a single variant,
    containing a single parameter, the dereference operator accesses
    this parameter.
`!`
  : Logical negation. On the boolean type, this flips between `true` and
    `false`. On integer types, this inverts the individual bits in the
    two's complement representation of the value.
`@` and `~`
  :  [Boxing](#box-types) operators. Allocate a box to hold the value
     they are applied to, and store the value in it. `@` creates a
     shared, reference-counted box, whereas `~` creates a unique box.

### Binary operator expressions

#### Arithmetic operators

Binary arithmetic expressions require both their operands to be of the
same type, and can be applied only to numeric types, with the
exception of `+`, which acts both as addition operator on numbers and
as concatenate operator on vectors and strings.

`+`
  : Addition and vector/string concatenation.
`-`
  : Subtraction.
`*`
  : Multiplication.
`/`
  : Division.
`%`
  : Remainder.

#### Bitwise operators

Bitwise operators apply only to integer types, and perform their
operation on the bits of the two's complement representation of the
values.

`&`
  : And.
`|`
  : Inclusive or.
`^`
  : Exclusive or.
`<<`
  : Logical left shift.
`>>`
  : Logical right shift.
`>>>`
  : Arithmetic right shift.

#### Lazy boolean operators

The operators `||` and `&&` may be applied to operands of boolean
type. The first performs the 'or' operation, and the second the 'and'
operation. They differ from `|` and `&` in that the right-hand operand
is only evaluated when the left-hand operand does not already
determine the outcome of the expression. That is, `||` only evaluates
it right-hand operand when the left-hand operand evaluates to `false`,
and `&&` only when it evaluates to `true`.

#### Comparison operators

`==`
  : Equal to.
`!=`
  : Unequal to.
`<`
  : Less than.
`>`
  : Greater than.
`<=`
  : Less than or equal.
`>=`
  : Greater than or equal.

The binary comparison operators can be applied to any two operands of
the same type, and produce a boolean value.

*TODO* details on how types are descended during comparison.

#### Type cast expressions

A type cast expression is denoted with the binary operator `as`.

Executing an `as` expression casts the value on the left-hand side to the type
on the right-hand side.

A numeric value can be cast to any numeric type.  A native pointer value can
be cast to or from any integral type or native pointer type.  Any other cast
is unsupported and will fail to compile.

An example of an `as` expression:

~~~~
fn avg(v: [float]) -> float {
  let sum: float = sum(v);
  let sz: float = std::vec::len(v) as float;
  ret sum / sz;
}
~~~~

A cast is a *trivial cast* iff the type of the casted expression and the
target type are identical after replacing all occurences of `int`, `uint`,
`float` with their machine type equivalents of the target architecture in both
types.


#### Binary move expressions

A _binary move experssion_ consists of an *lval* followed by a left-pointing
arrow (`<-`) and an *rval* expression.

Evaluating a move expression causes, as a side effect, the *rval* to be
*moved* into the *lval*. If the *rval* was itself an *lval*, it must be a
local variable, as it will be de-initialized in the process.

Evaluating a move expression does not effect reference counts nor does it
cause a deep copy of any unique structure pointed-to by the moved
*rval*. Instead, the move expression represents an indivisible *transfer of
ownership* from the right-hand-side to the left-hand-side of the
expression. No allocation or destruction is entailed.

An example of three different move expressions:

~~~~~~~~
x <- a;
x[i] <- b;
x.y <- c;
~~~~~~~~

#### Swap expressions

A _swap experssion_ consists of an *lval* followed by a bi-directional arrow
(`<->`) and another *lval* expression.

Evaluating a swap expression causes, as a side effect, the vales held in the
left-hand-side and right-hand-side *lvals* to be exchanged indivisibly.

Evaluating a move expression does not effect reference counts nor does it
cause a deep copy of any unique structure pointed-to by the moved
*rval*. Instead, the move expression represents an indivisible *exchange of
ownership* between the right-hand-side to the left-hand-side of the
expression. No allocation or destruction is entailed.

An example of three different swap expressions:

~~~~~~~~
x <-> a;
x[i] <-> b[i];
x.y <-> a.b;
~~~~~~~~


#### Assignment expressions

An _assignment expression_ consists of an *lval* expression followed by an
equals-sign (`=`) and an *rval* expression.

Evaluating an assignment expression is equivalent to evaluating a [binary move
expression](#binary-move-expressions) applied to a [unary copy
expression](unary-copy-expressions). For example, the following two
expressions have the same effect:

~~~~
x = y
x <- copy y
~~~~

The former is just more terse and familiar.

#### Operator-assignment expressions

The `+`, `-`, `*`, `/`, `%`, `&`, `|`, `^`, `<<`, `>>`, and `>>>`
operators may be composed with the `=` operator. The expression `lval
OP= val` is equivalent to `lval = lval OP val`. For example, `x = x +
1` may be written as `x += 1`.

#### Operator precedence

The precedence of Rust binary operators is ordered as follows, going
from strong to weak:

~~~~ {.precedence}
* / %
+ -
<< >> >>>
&
^ |
as
< > <= >=
== !=
&&
||
~~~~

### Unary copy expressions

A _unary copy expression_ consists of the unary `copy` operator applied to
some argument expression.

Evaluating a copy expression first evaluates the argument expression, then
performs a copy of the resulting value, allocating any memory necessary to
hold the new copy.

[Shared boxes](#shared-box-types) (type `@`) are, as usual, shallow-copied, as
they may be cyclic. [Unique boxes](unique-box-types), [vectors](#vector-types)
and similar unique types are deep-copied.

Since the binary [assignment operator](#assignment-operator) `=` performs a
copy implicitly, the unary copy operator is typically only used to cause an
argument to a function should be copied, and the copy passed by-value.

An example of a copy expression:

~~~~
fn mutate(vec: [mutable int]) {
   vec[0] = 10;
}

let v = [mutable 1,2,3];

mutate(copy v);   // Pass a copy

assert v[0] == 1; // Original was not modified
~~~~

### Unary move expressions

This is used to indicate that the referenced _lval_ must be moved out,
rather than copied, when evaluating this expression. It will only have
effect when the expression is _stored_ somewhere or passed to a
function that takes ownership of it.

~~~~
let x = ~10;
let y = [move x];
~~~~

Any access to `y` after applying the `move` operator to it is invalid,
since it is no longer initialized at that point.

### Call expressions

~~~~~~~~ {.abnf .gram}
expr_list : [ expr [ ',' expr ]* ] ? ;
paren_expr_list : '(' expr_list ')' ;
call_expr : expr paren_expr_list ;
~~~~~~~~

A _call expression_ invokes a function, providing a tuple of input slots
and a reference slot to serve as the function's output, bound to the
`lval` on the right hand side of the call. If the function eventually
returns, then the expression completes.

A call expression statically requires that the precondition declared in the
callee's signature is satisfied by the expression prestate. In this way,
typestates propagate through function boundaries. See [Ref.Typestate](#ref.typestate).

An example of a call expression:

~~~~
let x: int = add(1, 2);
~~~~


### Bind expressions

A _bind expression_ constructs a new function from an existing function.^[The
`bind` expression is analogous to the `bind` expression in the Sather
language.] The new function has zero or more of its arguments *bound* into a
new, hidden boxed tuple that holds the bindings. For each concrete argument
passed in the `bind` expression, the corresponding parameter in the existing
function is *omitted* as a parameter of the new function. For each argument
passed the placeholder symbol `_` in the `bind` expression, the corresponding
parameter of the existing function is *retained* as a parameter of the new
function.

Any subsequent invocation of the new function with residual arguments causes
invocation of the existing function with the combination of bound arguments
and residual arguments that was specified during the binding.

An example of a `bind` expression:

~~~~
fn add(x: int, y: int) -> int {
    ret x + y;
}
type single_param_fn = fn(int) -> int;

let add4: single_param_fn = bind add(4, _);

let add5: single_param_fn = bind add(_, 5);

assert (add(4,5) == add4(5));
assert (add(4,5) == add5(4));

~~~~

A `bind` expression generally stores a copy of the bound arguments in the
hidden, boxed tuple, owned by the resulting first-class function. For each
bound slot in the bound function's signature, space is allocated in the hidden
tuple and populated with a copy of the bound value.

A `bind` expression is an alternative way of constructing a shared function
closure; the [`fn@` expression](#shared-function-expression) form is another
way.

### Shared function expressions

*TODO*.

### Unique function expressions

*TODO*.

### While expressions

A `while` expression is a loop construct. A `while` loop may be either a
simple `while` or a `do`-`while` loop.

In the case of a simple `while`, the loop begins by evaluating the boolean
loop conditional expression. If the loop conditional expression evaluates to
`true`, the loop body block executes and control returns to the loop
conditional expression. If the loop conditional expression evaluates to
`false`, the `while` expression completes.

In the case of a `do`-`while`, the loop begins with an execution of the loop
body. After the loop body executes, it evaluates the loop conditional
expression. If it evaluates to `true`, control returns to the beginning of the
loop body. If it evaluates to `false`, control exits the loop.

An example of a simple `while` expression:

~~~~
while (i < 10) {
    print("hello\n");
    i = i + 1;
}
~~~~

An example of a `do`-`while` expression:

~~~~
do {
    print("hello\n");
    i = i + 1;
} while (i < 10);
~~~~


### Break expressions

Executing a `break` expression immediately terminates the innermost loop
enclosing it. It is only permitted in the body of a loop.

### Continue expressions

Evaluating a `cont` expression immediately terminates the current iteration of
the innermost loop enclosing it, returning control to the loop *head*. In the
case of a `while` loop, the head is the conditional expression controlling the
loop. In the case of a `for` loop, the head is the vector-element increment
controlling the loop.

A `cont` expression is only permitted in the body of a loop.


### For expressions

A _for loop_ is controlled by a vector or string. The for loop bounds-checks
the underlying sequence *once* when initiating the loop, then repeatedly
executes the loop body with the loop variable referencing the successive
elements of the underlying sequence, one iteration per sequence element.

An example a for loop:

~~~~
let v: [foo] = [a, b, c];

for e: foo in v {
    bar(e);
}
~~~~


### If expressions

An `if` expression is a conditional branch in program control. The form of
an `if` expression is a condition expression, followed by a consequent
block, any number of `else if` conditions and blocks, and an optional
trailing `else` block. The condition expressions must have type
`bool`. If a condition expression evaluates to `true`, the
consequent block is executed and any subsequent `else if` or `else`
block is skipped. If a condition expression evaluates to `false`, the
consequent block is skipped and any subsequent `else if` condition is
evaluated. If all `if` and `else if` conditions evaluate to `false`
then any `else` block is executed.


### Alternative expressions

An `alt` expression branches on a *pattern*. The exact form of matching that
occurs depends on the pattern. Patterns consist of some combination of
literals, destructured tag constructors, records and tuples, variable binding
specifications and placeholders (`_`). An `alt` expression has a *head
expression*, which is the value to compare to the patterns. The type of the
patterns must equal the type of the head expression.

To execute an `alt` expression, first the head expression is evaluated, then
its value is sequentially compared to the patterns in the arms until a match
is found. The first arm with a matching pattern is chosen as the branch target
of the `alt`, any variables bound by the pattern are assigned to local slots
in the arm's block, and control enters the block.

An example of an `alt` expression:


~~~~
tag list<X> { nil; cons(X, @list<X>); }

let x: list<int> = cons(10, @cons(11, @nil));

alt x {
    cons(a, @cons(b, _)) {
        process_pair(a,b);
    }
    cons(10, _) {
        process_ten();
    }
    nil {
        ret;
    }
    _ {
        fail;
    }
}
~~~~

Records can also be pattern-matched and their fields bound to variables.
When matching fields of a record, the fields being matched are specified
first, then a placeholder (`_`) represents the remaining fields.

~~~~
fn main() {
    let r = {
        player: "ralph",
        stats: load_stats(),
        options: {
            choose: true,
            size: "small"
        }
    };

    alt r {
      {options: {choose: true, _}, _} {
        choose_player(r)
      }
      {player: p, options: {size: "small", _}, _} {
        log(info, p + " is small");
      }
      _ {
        next_player();
      }
    }
}
~~~~

Multiple alternative patterns may be joined with the `|` operator.  A
range of values may be specified with `to`.  For example:

~~~~
let message = alt x {
  0 | 1  { "not many" }
  2 to 9 { "a few" }
  _      { "lots" }
}
~~~~

Finally, alt patterns can accept *pattern guards* to further refine the
criteria for matching a case. Pattern guards appear after the pattern and
consist of a bool-typed expression following the `if` keyword. A pattern
guard may refer to the variables bound within the pattern they follow.

~~~~
let message = alt maybe_digit {
  some(x) if x < 10 { process_digit(x) }
  some(x) { process_other(x) }
}
~~~~


### Fail expressions

Evaluating a `fail` expression causes a task to enter the *failing* state. In
the *failing* state, a task unwinds its stack, destroying all frames and
freeing all resources until it reaches its entry frame, at which point it
halts execution in the *dead* state.

### Note expressions

**Note: Note expressions are not yet supported by the compiler.**

A `note` expression has no effect during normal execution. The purpose of a
`note` expression is to provide additional diagnostic information to the
logging subsystem during task failure. See [log
expressions](#log-expressions). Using `note` expressions, normal diagnostic
logging can be kept relatively sparse, while still providing verbose
diagnostic "back-traces" when a task fails.

When a task is failing, control frames *unwind* from the innermost frame to
the outermost, and from the innermost lexical block within an unwinding frame
to the outermost. When unwinding a lexical block, the runtime processes all
the `note` expressions in the block sequentially, from the first expression of
the block to the last.  During processing, a `note` expression has equivalent
meaning to a `log` expression: it causes the runtime to append the argument of
the `note` to the internal logging diagnostic buffer.

An example of a `note` expression:

~~~~
fn read_file_lines(path: str) -> [str] {
    note path;
    let r: [str];
    let f: file = open_read(path);
    lines(f) {|s|
        r += [s];
    }
    ret r;
}
~~~~

In this example, if the task fails while attempting to open or read a file,
the runtime will log the path name that was being read. If the function
completes normally, the runtime will not log the path.

A value that is marked by a `note` expression is *not* copied aside
when control passes through the `note`. In other words, if a `note`
expression notes a particular `lval`, and code after the `note`
mutates that slot, and then a subsequent failure occurs, the *mutated*
value will be logged during unwinding, *not* the original value that was
denoted by the `lval` at the moment control passed through the `note`
expression.

### Return expressions

Return expressions are denoted with the keyword `ret`. Evaluating a `ret`
expression^[footnote{A `ret` expression is analogous to a `return` expression
in the C family.] moves its argument into the output slot of the current
function, destroys the current function activation frame, and transfers
control to the caller frame.

An example of a `ret` expression:

~~~~
fn max(a: int, b: int) -> int {
   if a > b {
      ret a;
   }
   ret b;
}
~~~~

### Log expressions

Evaluating a `log` expression may, depending on runtime configuration, cause a
value to be appended to an internal diagnostic logging buffer provided by the
runtime or emitted to a system console. Log expressions are enabled or
disabled dynamically at run-time on a per-task and per-item basis. See
[logging system](#logging-system).

Each `log` expression must be provided with a *level* argument in
addition to the value to log. The logging level is a `u32` value, where
lower levels indicate more-urgent levels of logging. By default, the lowest
four logging levels (`0_u32 ... 3_u32`) are predefined as the constants
`error`, `warn`, `info` and `debug` in the `core` library.

Additionally, the macros `#error`, `#warn`, `#info` and `#debug` are defined
in the default syntax-extension namespace. These expand into calls to the
logging facility composed with calls to the `#fmt` string formatting
syntax-extension.

The following examples all produce the same output, logged at the `error`
logging level:

~~~~
// Full version, logging a value.
log(core::error, "file not found: " + filename);

// Log-level abbreviated, since core::* is imported by default.
log(error, "file not found: " + filename);

// Formatting the message using a format-string and #fmt
log(error, #fmt("file not found: %s", filename));

// Using the #error macro, that expands to the previous call.
#error("file not found: %s", filename);
~~~~

A `log` expression is *not evaluated* when logging at the specified
logging-level, module or task is disabled at runtime. This makes inactive
`log` expressions very cheap; they should be used extensively in Rust
code, as diagnostic aids, as they add little overhead beyond a single
integer-compare and branch at runtime.

Logging is presently implemented as a language built-in feature, as it makes
use of compiler-provided logic for allocating the associated per-module
logging-control structures visible to the runtime, and lazily evaluating
arguments. In the future, as more of the supporting compiler-provided logic is
moved into libraries, logging is likely to move to a component of the core
library. It is best to use the macro forms of logging (*#error*,
*#debug*, etc.) to minimize disruption to code using the logging facility
when it is changed.


### Check expressions

A `check` expression connects dynamic assertions made at run-time to the
static [typestate system](#typestate-system). A `check` expression takes a
constraint to check at run-time. If the constraint holds at run-time, control
passes through the `check` and on to the next expression in the enclosing
block. If the condition fails to hold at run-time, the `check` expression
behaves as a `fail` expression.

The typestate algorithm is built around `check` expressions, and in particular
the fact that control *will not pass* a check expression with a condition that
fails to hold. The typestate algorithm can therefore assume that the (static)
postcondition of a `check` expression includes the checked constraint
itself. From there, the typestate algorithm can perform dataflow calculations
on subsequent expressions, propagating [conditions](#conditions) forward and
statically comparing implied states and their specifications.

~~~~~~~~
pure fn even(x: int) -> bool {
    ret x & 1 == 0;
}

fn print_even(x: int) : even(x) {
    print(x);
}

fn test() {
    let y: int = 8;

    // Cannot call print_even(y) here.

    check even(y);

    // Can call print_even(y) here, since even(y) now holds.
    print_even(y);
}
~~~~~~~~

### Prove expressions

A `prove` expression has no run-time effect. Its purpose is to statically
check (and document) that its argument constraint holds at its expression
entry point. If its argument typestate does not hold, under the typestate
algorithm, the program containing it will fail to compile.

### Claim expressions

A `claim` expression is an unsafe variant on a `check` expression that is not
actually checked at runtime. Thus, using a `claim` implies a proof obligation
to ensure---without compiler assistance---that an assertion always holds.

Setting a runtime flag can turn all `claim` expressions into `check`
expressions in a compiled Rust program, but the default is to not check the
assertion contained in a `claim`. The idea behind `claim` is that performance
profiling might identify a few bottlenecks in the code where actually checking
a given callee's predicate is too expensive; `claim` allows the code to
typecheck without removing the predicate check at every other call site.



### If-Check expressions

An `if check` expression combines a `if` expression and a `check`
expression in an indivisible unit that can be used to build more complex
conditional control-flow than the `check` expression affords.

In fact, `if check` is a "more primitive" expression than `check`;
instances of the latter can be rewritten as instances of the former. The
following two examples are equivalent:

Example using `check`:

~~~~
check even(x);
print_even(x);
~~~~

Equivalent example using `if check`:

~~~~
if check even(x) {
    print_even(x);
} else {
    fail;
}
~~~~

### Assert expressions

An `assert` expression is similar to a `check` expression, except
the condition may be any boolean-typed expression, and the compiler makes no
use of the knowledge that the condition holds if the program continues to
execute after the `assert`.


### Syntax extension expressions

~~~~~~~~ {.abnf .gram}
syntax_ext_expr : '#' ident paren_expr_list ? brace_match ? ;
~~~~~~~~

Rust provides a notation for _syntax extension_. The notation for invoking
a syntax extension is a marked syntactic form that can appear as an expression
in the body of a Rust program.

After parsing, a syntax-extension invocation is expanded into a Rust
expression. The name of the extension determines the translation performed. In
future versions of Rust, user-provided syntax extensions aside from macros
will be provided via external crates.

At present, only a set of built-in syntax extensions, as well as macros
introduced inline in source code using the `macro` extension, may be used. The
current built-in syntax extensions are:


* `fmt` expands into code to produce a formatted string, similar to 
      `printf` from C.
* `env` expands into a string literal containing the value of that
      environment variable at compile-time.
* `concat_idents` expands into an identifier which is the 
      concatenation of its arguments.
* `ident_to_str` expands into a string literal containing the name of
      its argument (which must be a literal).
* `log_syntax` causes the compiler to pretty-print its arguments.


Finally, `macro` is used to define a new macro. A macro can abstract over
second-class Rust concepts that are present in syntax. The arguments to
`macro` are pairs (two-element vectors). The pairs consist of an invocation
and the syntax to expand into. An example:

~~~~~~~~
#macro([#apply[fn, [args, ...]], fn(args, ...)]);
~~~~~~~~

In this case, the invocation `#apply[sum, 5, 8, 6]` expands to
`sum(5,8,6)`. If `...` follows an expression (which need not be as
simple as a single identifier) in the input syntax, the matcher will expect an
arbitrary number of occurrences of the thing preceding it, and bind syntax to
the identifiers it contains. If it follows an expression in the output syntax,
it will transcribe that expression repeatedly, according to the identifiers
(bound to syntax) that it contains.

The behaviour of `...` is known as Macro By Example. It allows you to
write a macro with arbitrary repetition by specifying only one case of that
repetition, and following it by `...`, both where the repeated input is
matched, and where the repeated output must be transcribed. A more
sophisticated example:


~~~~~~~~
#macro([#zip_literals[[x, ...], [y, ...]), [[x, y], ...]]);
#macro([#unzip_literals[[x, y], ...], [[x, ...], [y, ...]]]);
~~~~~~~~

In this case, `#zip_literals[[1,2,3], [1,2,3]]` expands to
`[[1,1],[2,2],[3,3]]`, and `#unzip_literals[[1,1], [2,2], [3,3]]`
expands to `[[1,2,3],[1,2,3]]`.

Macro expansion takes place outside-in: that is,
`#unzip_literals[#zip_literals[[1,2,3],[1,2,3]]]` will fail because
`unzip_literals` expects a list, not a macro invocation, as an argument.

The macro system currently has some limitations. It's not possible to
destructure anything other than vector literals (therefore, the arguments to
complicated macros will tend to be an ocean of square brackets). Macro
invocations and `...` can only appear in expression positions. Finally,
macro expansion is currently unhygienic. That is, name collisions between
macro-generated and user-written code can cause unintentional capture.

Future versions of Rust will address these issues.


# Types and typestates

## Type system

Every slot and value in a Rust program has a type. The _type_ of a *value*
defines the interpretation of the memory holding it. The type of a *slot* may
also include [constraints](#constrained-types).

Built-in types and type-constructors are tightly integrated into the language,
in nontrivial ways that are not possible to emulate in user-defined
types. User-defined types have limited capabilities. In addition, every
built-in type or type-constructor name is reserved as a *keyword* in Rust;
they cannot be used as user-defined identifiers in any context.

### Primitive types

The primitive types are the following:

* The "nil" type `()`, having the single "nil" value `()`.^[The "nil" value
  `()` is *not* a sentinel "null pointer" value for reference slots; the "nil"
  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.]
* The boolean type `bool` with values `true` and `false`.
* The machine types.
* The machine-dependent integer and floating-point types.

#### Machine types

The machine types are the following:


* The unsigned word types `u8`, `u16`, `u32` and `u64`, with values drawn from
  the integer intervals $[0, 2^8 - 1]$, $[0, 2^16 - 1]$, $[0, 2^32 - 1]$ and
  $[0, 2^64 - 1]$ respectively.

* The signed two's complement word types `i8`, `i16`, `i32` and `i64`, with
  values drawn from the integer intervals $[-(2^7), 2^7 - 1]$,
  $[-(2^15), 2^15 - 1]$, $[-(2^31), 2^31 - 1]$, $[-(2^63), 2^63 - 1]$
  respectively.

* The IEEE 754-2008 `binary32` and `binary64` floating-point types: `f32` and
  `f64`, respectively.

#### Machine-dependent integer types

The Rust type `uint`^[A Rust `uint` is analogous to a C99 `uintptr_t`.] is an
unsigned integer type with with target-machine-dependent size. Its size, in
bits, is equal to the number of bits required to hold any memory address on
the target machine.

The Rust type `int`^[A Rust `int` is analogous to a C99 `intptr_t`.] is a
two's complement signed integer type with target-machine-dependent size. Its
size, in bits, is equal to the size of the rust type `uint` on the same target
machine.


#### Machine-dependent floating point type

The Rust type `float` is a machine-specific type equal to one of the supported
Rust floating-point machine types (`f32` or `f64`). It is the largest
floating-point type that is directly supported by hardware on the target
machine, or if the target machine has no floating-point hardware support, the
largest floating-point type supported by the software floating-point library
used to support the other floating-point machine types.

Note that due to the preference for hardware-supported floating-point, the
type `float` may not be equal to the largest *supported* floating-point type.


### Textual types

The types `char` and `str` hold textual data.

A value of type `char` is a Unicode character, represented as a 32-bit
unsigned word holding a UCS-4 codepoint.

A value of type `str` is a Unicode string, represented as a vector of 8-bit
unsigned bytes holding a sequence of UTF-8 codepoints.



### Record types

The record type-constructor forms a new heterogeneous product of values.^[The
record type-constructor is analogous to the `struct` type-constructor in the
Algol/C family, the *record* types of the ML family, or the *structure* types
of the Lisp family.] Fields of a record type are accessed by name and are
arranged in memory in the order specified by the record type.

An example of a record type and its use:

~~~~
type point = {x: int, y: int};
let p: point = {x: 10, y: 11};
let px: int = p.x;
~~~~

### Tuple types

The tuple type-constructor forms a new heterogeneous product of values similar
to the record type-constructor. The differences are as follows:

* tuple elements cannot be mutable, unlike record fields
* tuple elements are not named and can be accessed only by pattern-matching

Tuple types and values are denoted by listing the types or values of their
elements, respectively, in a parenthesized, comma-separated
list. Single-element tuples are not legal; all tuples have two or more values.

The members of a tuple are laid out in memory contiguously, like a record, in
order specified by the tuple type.

An example of a tuple type and its use:

~~~~
type pair = (int,str);
let p: pair = (10,"hello");
let (a, b) = p;
assert (b == "world");
~~~~

### Vector types

The vector type-constructor represents a homogeneous array of values of a
given type. A vector has a fixed size. The kind of a vector type depends on
the kind of its member type, as with other simple structural types.

An example of a vector type and its use:

~~~~
let v: [int] = [7, 5, 3];
let i: int = v[2];
assert (i == 3);
~~~~

Vectors always *allocate* a storage region sufficient to store the first power
of two worth of elements greater than or equal to the size of the vector. This
behaviour supports idiomatic in-place "growth" of a mutable slot holding a
vector:


~~~~
let v: mutable [int] = [1, 2, 3];
v += [4, 5, 6];
~~~~

Normal vector concatenation causes the allocation of a fresh vector to hold
the result; in this case, however, the slot holding the vector recycles the
underlying storage in-place (since the reference-count of the underlying
storage is equal to 1).

All accessible elements of a vector are always initialized, and access to a
vector is always bounds-checked.


### Enumerated types

An *enumerated type* is a nominal, heterogeneous disjoint union type.^[The
`enum` type is analogous to a `data` constructor declaration in ML or a *pick
ADT* in Limbo.} An [`enum` *item*](#enumerations) consists of a number of
*constructors*, each of which is independently named and takes an optional
tuple of arguments.

Enumerated types cannot be denoted *structurally* as types, but must be
denoted by named reference to an [*enumeration* item](#enumerations).

### Box types

Box types are represented as pointers. There are three flavours of
pointers:

Shared boxes (`@`)
  : These are reference-counted boxes. Their type is written
    `@content`, for example `@int` means a shared box containing an
    integer. Copying a value of such a type means copying the pointer
    and increasing the reference count.

Unique boxes (`~`)
  : Unique boxes have only a single owner, and are freed when their
    owner releases them. They are written `~content`. Copying a
    unique box involves copying the contents into a new box.

Unsafe pointers (`*`)
  : Unsafe pointers are pointers without safety guarantees or
    language-enforced semantics. Their type is written `*content`.
    They can be copied and dropped freely. Dereferencing an unsafe
    pointer is part of the unsafe sub-dialect of Rust.

### Function types

The function type-constructor `fn` forms new function types. A function type
consists of a sequence of input slots, an optional set of [input
constraints](#input-constraints) and an output slot. See
[Ref.Item.Fn](#ref.item.fn).

An example of a `fn` type:

~~~~~~~~
fn add(x: int, y: int) -> int {
  ret x + y;
}

let int x = add(5,7);

type binop = fn(int,int) -> int;
let bo: binop = add;
x = bo(5,7);
~~~~~~~~

## Type kinds

Types in Rust are categorized into three kinds, based on whether they
allow copying of their values, and sending to different tasks. The
kinds are:

Sendable
  : Values with a sendable type can be safely sent to another task.
    This kind includes scalars, unique pointers, unique closures, and
    structural types containing only other sendable types.
Copyable
  : This kind includes all types that can be copied. All types with
    sendable kind are copyable, as are shared boxes, shared closures,
    interface types, and structural types built out of these.
Noncopyable
  : [Resource](#resources) types, and every type that includes a
    resource without storing it in a shared box, may not be copied.
    Types of sendable or copyable type can always be used in places
    where a noncopyable type is expected, so in effect this kind
    includes all types.

These form a hierarchy. The noncopyable kind is the widest, including
all types in the language. The copyable kind is a subset of that, and
the sendable kind is a subset of the copyable kind.

Any operation that causes a value to be copied requires the type of
that value to be of copyable kind. Type parameter types are assumed to
be noncopyable, unless one of the special bounds `send` or `copy` is
declared for it. For example, this is not a valid program:

~~~~
fn box<T>(x: T) -> @T { @x }
~~~~

Putting `x` into a shared box involves copying, and the `T` parameter
is assumed to be noncopyable. To change that, a bound is declared:

~~~~
fn box<T: copy>(x: T) -> @T { @x }
~~~~

Calling this second version of `box` on a noncopyable type is not
allowed. When instantiating a type parameter, the kind bounds on the
parameter are checked to be the same or narrower than the kind of the
type that it is instantiated with.

Sending operations are not part of the Rust language, but are
implemented in the library. Generic functions that send values bound
the kind of these values to sendable.



## Typestate system


Rust programs have a static semantics that determine the types of values
produced by each expression, as well as the *predicates* that hold over
slots in the environment at each point in time during execution.

The latter semantics -- the dataflow analysis of predicates holding over slots
-- is called the *typestate* system.

### Points

Control flows from statement to statement in a block, and through the
evaluation of each expression, from one sub-expression to another. This
sequential control flow is specified as a set of _points_, each of which
has a set of points before and after it in the implied control flow.

For example, this code:

~~~~~~~~
 s = "hello, world";
 print(s);
~~~~~~~~

Consists of 2 statements, 3 expressions and 12 points:


* the point before the first statement
* the point before evaluating the static initializer `"hello, world"`
* the point after evaluating the static initializer `"hello, world"`
* the point after the first statement
* the point before the second statement
* the point before evaluating the function value `print`
* the point after evaluating the function value `print`
* the point before evaluating the arguments to `print`
* the point before evaluating the symbol `s`
* the point after evaluating the symbol `s`
* the point after evaluating the arguments to `print`
* the point after the second statement


Whereas this code:


~~~~~~~~
 print(x() + y());
~~~~~~~~

Consists of 1 statement, 7 expressions and 14 points:


* the point before the statement
* the point before evaluating the function value `print`
* the point after evaluating the function value `print`
* the point before evaluating the arguments to `print`
* the point before evaluating the arguments to `+`
* the point before evaluating the function value `x`
* the point after evaluating the function value `x`
* the point before evaluating the arguments to `x`
* the point after evaluating the arguments to `x`
* the point before evaluating the function value `y`
* the point after evaluating the function value `y`
* the point before evaluating the arguments to `y`
* the point after evaluating the arguments to `y`
* the point after evaluating the arguments to `+`
* the point after evaluating the arguments to `print`


The typestate system reasons over points, rather than statements or
expressions. This may seem counter-intuitive, but points are the more
primitive concept. Another way of thinking about a point is as a set of
*instants in time* at which the state of a task is fixed. By contrast, a
statement or expression represents a *duration in time*, during which the
state of the task changes. The typestate system is concerned with constraining
the possible states of a task's memory at *instants*; it is meaningless to
speak of the state of a task's memory "at" a statement or expression, as each
statement or expression is likely to change the contents of memory.


### Control flow graph

Each *point* can be considered a vertex in a directed *graph*. Each
kind of expression or statement implies a number of points *and edges* in
this graph. The edges connect the points within each statement or expression,
as well as between those points and those of nearby statements and expressions
in the program. The edges between points represent *possible* indivisible
control transfers that might occur during execution.

This implicit graph is called the _control-flow graph_, or _CFG_.


### Constraints

A [_predicate_](#predicate-functions) is a pure boolean function declared with
the keywords `pure fn`.

A _constraint_ is a predicate applied to specific slots.

For example, consider the following code:

~~~~~~~~
pure fn is_less_than(int a, int b) -> bool {
     ret a < b;
}

fn test() {
   let x: int = 10;
   let y: int = 20;
   check is_less_than(x,y);
}
~~~~~~~~

This example defines the predicate `is_less_than`, and applies it to the slots
`x` and `y`. The constraint being checked on the third line of the function is
`is_less_than(x,y)`.

Predicates can only apply to slots holding immutable values. The slots a
predicate applies to can themselves be mutable, but the types of values held
in those slots must be immutable.

### Conditions

A _condition_ is a set of zero or more constraints.

Each *point* has an associated *condition*:

* The _precondition_ of a statement or expression is the condition required at
in the point before it.
* The _postcondition_ of a statement or expression is the condition enforced
in the point after it.

Any constraint present in the precondition and *absent* in the postcondition
is considered to be *dropped* by the statement or expression.


### Calculated typestates

The typestate checking system *calculates* an additional condition for each
point called its _typestate_. For a given statement or expression, we call the
two typestates associated with its two points the prestate and a poststate.

* The _prestate_ of a statement or expression is the typestate of the
point before it.
* The _poststate_ of a statement or expression is the typestate of the
point after it.

A _typestate_ is a condition that has _been determined by the typestate
algorithm_ to hold at a point. This is a subtle but important point to
understand: preconditions and postconditions are *inputs* to the typestate
algorithm; prestates and poststates are *outputs* from the typestate
algorithm.

The typestate algorithm analyses the preconditions and postconditions of every
statement and expression in a block, and computes a condition for each
typestate. Specifically:


* Initially, every typestate is empty.
* Each statement or expression's poststate is given the union of the its
prestate, precondition, and postcondition.
* Each statement or expression's poststate has the difference between its
precondition and postcondition removed.
* Each statement or expression's prestate is given the intersection of the
poststates of every predecessor point in the CFG.
* The previous three steps are repeated until no typestates in the
block change.

The typestate algorithm is a very conventional dataflow calculation, and can
be performed using bit-set operations, with one bit per predicate and one
bit-set per condition.

After the typestates of a block are computed, the typestate algorithm checks
that every constraint in the precondition of a statement is satisfied by its
prestate. If any preconditions are not satisfied, the mismatch is considered a
static (compile-time) error.


### Typestate checks

The key mechanism that connects run-time semantics and compile-time analysis
of typestates is the use of [`check` expressions](#check-expressions). A
`check` expression guarantees that *if* control were to proceed past it, the
predicate associated with the `check` would have succeeded, so the constraint
being checked *statically* holds in subsequent points.^[A `check` expression
is similar to an `assert` call in a C program, with the significant difference
that the Rust compiler *tracks* the constraint that each `check` expression
enforces. Naturally, `check` expressions cannot be omitted from a "production
build" of a Rust program the same way `asserts` are frequently disabled in
deployed C programs.}

It is important to understand that the typestate system has *no insight* into
the meaning of a particular predicate. Predicates and constraints are not
evaluated in any way at compile time. Predicates are treated as specific (but
unknown) functions applied to specific (also unknown) slots. All the typestate
system does is track which of those predicates -- whatever they calculate --
*must have been checked already* in order for program control to reach a
particular point in the CFG. The fundamental building block, therefore, is the
`check` statement, which tells the typestate system "if control passes this
point, the checked predicate holds".

From this building block, constraints can be propagated to function signatures
and constrained types, and the responsibility to `check` a constraint
pushed further and further away from the site at which the program requires it
to hold in order to execute properly.



# Memory and concurrency models

Rust has a memory model centered around concurrently-executing _tasks_. Thus
its memory model and its concurrency model are best discussed simultaneously,
as parts of each only make sense when considered from the perspective of the
other.

When reading about the memory model, keep in mind that it is partitioned in
order to support tasks; and when reading about tasks, keep in mind that their
isolation and communication mechanisms are only possible due to the ownership
and lifetime semantics of the memory model.

## Memory model

A Rust [task](#tasks)'s memory consists of a static set of *items*, a set of
tasks each with its own *stack*, and a *heap*. Immutable portions of the heap
may be shared between tasks, mutable portions may not.

Allocations in the stack consist of *slots*, and allocations in the heap
consist of *boxes*.


### Memory allocation and lifetime

The _items_ of a program are those functions, modules and types
that have their value calculated at compile-time and stored uniquely in the
memory image of the rust process. Items are neither dynamically allocated nor
freed.

A task's _stack_ consists of activation frames automatically allocated on
entry to each function as the task executes. A stack allocation is reclaimed
when control leaves the frame containing it.

The _heap_ is a general term that describes two separate sets of boxes:
shared boxes -- which may be subject to garbage collection -- and unique
boxes.  The lifetime of an allocation in the heap depends on the lifetime of
the box values pointing to it. Since box values may themselves be passed in
and out of frames, or stored in the heap, heap allocations may outlive the
frame they are allocated within.


### Memory ownership

A task owns all memory it can *safely* reach through local variables,
shared or unique boxes, and/or references. Sharing memory between tasks can
only be accomplished using *unsafe* constructs, such as raw pointer
operations or calling C code.

When a task sends a value satisfying the `send` interface over a channel, it
loses ownership of the value sent and can no longer refer to it. This is
statically guaranteed by the combined use of "move semantics" and the
compiler-checked _meaning_ of the `send` interface: it is only instantiated
for (transitively) unique kinds of data constructor and pointers, never shared
pointers.

When a stack frame is exited, its local allocations are all released, and its
references to boxes (both shared and owned) are dropped.

A shared box may (in the case of a recursive, mutable shared type) be cyclic;
in this case the release of memory inside the shared structure may be deferred
until task-local garbage collection can reclaim it. Code can ensure no such
delayed deallocation occurs by restricting itself to unique boxes and similar
unshared kinds of data.

When a task finishes, its stack is necessarily empty and it therefore has no
references to any boxes; the remainder of its heap is immediately freed.


### Memory slots

A task's stack contains slots.

A _slot_ is a component of a stack frame. A slot is either *local* or
a *reference*.

A _local_ slot (or *stack-local* allocation) holds a value directly,
allocated within the stack's memory. The value is a part of the stack frame.

A _reference_ references a value outside the frame. It may refer to a
value allocated in another frame *or* a boxed value in the heap. The
reference-formation rules ensure that the referent will outlive the reference.

Local slots are always implicitly mutable.

Local slots are not initialized when allocated; the entire frame worth of
local slots are allocated at once, on frame-entry, in an uninitialized
state. Subsequent statements within a function may or may not initialize the
local slots. Local slots can be used only after they have been initialized;
this condition is guaranteed by the typestate system.

References are created for function arguments. If the compiler can not prove
that the referred-to value will outlive the reference, it will try to set
aside a copy of that value to refer to. If this is not semantically safe (for
example, if the referred-to value contains mutable fields), it will reject the
program. If the compiler deems copying the value expensive, it will warn.

A function can be declared to take an argument by mutable reference. This
allows the function to write to the slot that the reference refers to.

An example function that accepts an value by mutable reference:

~~~~~~~~
fn incr(&i: int) {
    i = i + 1;
}
~~~~~~~~

### Memory boxes

A _box_ is a reference to a heap allocation holding another value. There
are two kinds of boxes: *shared boxes* and *unique boxes*.

A _shared box_ type or value is constructed by the prefix *at* sigil `@`.

A _unique box_ type or value is constructed by the prefix *tilde* sigil `~`.

Multiple shared box values can point to the same heap allocation; copying a
shared box value makes a shallow copy of the pointer (optionally incrementing
a reference count, if the shared box is implemented through
reference-counting).

Unique box values exist in 1:1 correspondence with their heap allocation;
copying a unique box value makes a deep copy of the heap allocation and
produces a pointer to the new allocation.

An example of constructing one shared box type and value, and one unique box
type and value:

~~~~~~~~
let x: @int = @10;
let x: ~int = ~10;
~~~~~~~~

Some operations implicitly dereference boxes. Examples of such @dfn{implicit
dereference} operations are:

* arithmetic operators (`x + y - z`)
* field selection (`x.y.z`)


An example of an implicit-dereference operation performed on box values:

~~~~~~~~
let x: @int = @10;
let y: @int = @12;
assert (x + y == 22);
~~~~~~~~

Other operations act on box values as single-word-sized address values. For
these operations, to access the value held in the box requires an explicit
dereference of the box value. Explicitly dereferencing a box is indicated with
the unary *star* operator `*`. Examples of such @dfn{explicit
dereference} operations are:

* copying box values (`x = y`)
* passing box values to functions (`f(x,y)`)


An example of an explicit-dereference operation performed on box values:

~~~~~~~~
fn takes_boxed(b: @int) {
}

fn takes_unboxed(b: int) {
}

fn main() {
    let x: @int = @10;
    takes_boxed(x);
    takes_unboxed(*x);
}
~~~~~~~~

## Tasks

An executing Rust program consists of a tree of tasks. A Rust _task_
consists of an entry function, a stack, a set of outgoing communication
channels and incoming communication ports, and ownership of some portion of
the heap of a single operating-system process.

Multiple Rust tasks may coexist in a single operating-system process. The
runtime scheduler maps tasks to a certain number of operating-system threads;
by default a number of threads is used based on the number of concurrent
physical CPUs detected at startup, but this can be changed dynamically at
runtime. When the number of tasks exceeds the number of threads -- which is
quite possible -- the tasks are multiplexed onto the threads ^[This is an M:N
scheduler, which is known to give suboptimal results for CPU-bound concurrency
problems. In such cases, running with the same number of threads as tasks can
give better results. The M:N scheduling in Rust exists to support very large
numbers of tasks in contexts where threads are too resource-intensive to use
in a similar volume. The cost of threads varies substantially per operating
system, and is sometimes quite low, so this flexibility is not always worth
exploiting.]


### Communication between tasks

With the exception of *unsafe* blocks, Rust tasks are isolated from
interfering with one another's memory directly. Instead of manipulating shared
storage, Rust tasks communicate with one another using a typed, asynchronous,
simplex message-passing system.

A _port_ is a communication endpoint that can *receive* messages. Ports
receive messages from channels.

A _channel_ is a communication endpoint that can *send* messages. Channels
send messages to ports.

Each port is implicitly boxed and mutable; as such a port has a unique
per-task identity and cannot be replicated or transmitted. If a port value is
copied, both copies refer to the *same* port. New ports can be
constructed dynamically and stored in data structures.

Each channel is bound to a port when the channel is constructed, so the
destination port for a channel must exist before the channel itself. A channel
cannot be rebound to a different port from the one it was constructed with.

Channels are weak: a channel does not keep the port it is bound to
alive. Ports are owned by their allocating task and cannot be sent over
channels; if a task dies its ports die with it, and all channels bound to
those ports no longer function. Messages sent to a channel connected to a dead
port will be dropped.

Channels are immutable types with meaning known to the runtime; channels can
be sent over channels.

Many channels can be bound to the same port, but each channel is bound to a
single port. In other words, channels and ports exist in an N:1 relationship,
N channels to 1 port. ^[It may help to remember nautical terminology
when differentiating channels from ports.  Many different waterways --
channels -- may lead to the same port.}

Each port and channel can carry only one type of message. The message type is
encoded as a parameter of the channel or port type. The message type of a
channel is equal to the message type of the port it is bound to. The types of
messages must satisfy the `send` built-in interface.

Messages are generally sent asynchronously, with optional rate-limiting on the
transmit side. A channel contains a message queue and asynchronously sending a
message merely inserts it into the sending channel's queue; message receipt is
the responsibility of the receiving task.

Messages are sent on channels and received on ports using standard library
functions.


### Task lifecycle

The _lifecycle_ of a task consists of a finite set of states and events
that cause transitions between the states. The lifecycle states of a task are:

* running
* blocked
* failing
* dead

A task begins its lifecycle -- once it has been spawned -- in the *running*
state. In this state it executes the statements of its entry function, and any
functions called by the entry function.

A task may transition from the *running* state to the *blocked* state any time
it makes a blocking recieve call on a port, or attempts a rate-limited
blocking send on a channel. When the communication expression can be completed
-- when a message arrives at a sender, or a queue drains sufficiently to
complete a rate-limited send -- then the blocked task will unblock and
transition back to *running*.

A task may transition to the *failing* state at any time, due being
killed by some external event or internally, from the evaluation of a
`fail` expression. Once *failing*, a task unwinds its stack and
transitions to the *dead* state. Unwinding the stack of a task is done by
the task itself, on its own control stack. If a value with a destructor is
freed during unwinding, the code for the destructor is run, also on the task's
control stack. Running the destructor code causes a temporary transition to a
*running* state, and allows the destructor code to cause any subsequent
state transitions.  The original task of unwinding and failing thereby may
suspend temporarily, and may involve (recursive) unwinding of the stack of a
failed destructor. Nonetheless, the outermost unwinding activity will continue
until the stack is unwound and the task transitions to the *dead*
state. There is no way to "recover" from task failure.  Once a task has
temporarily suspended its unwinding in the *failing* state, failure
occurring from within this destructor results in *hard* failure.  The
unwinding procedure of hard failure frees resources but does not execute
destructors.  The original (soft) failure is still resumed at the point where
it was temporarily suspended.

A task in the *dead* state cannot transition to other states; it exists
only to have its termination status inspected by other tasks, and/or to await
reclamation when the last reference to it drops.


### Task scheduling

The currently scheduled task is given a finite *time slice* in which to
execute, after which it is *descheduled* at a loop-edge or similar
preemption point, and another task within is scheduled, pseudo-randomly.

An executing task can yield control at any time, by making a library call to
`std::task::yield`, which deschedules it immediately. Entering any other
non-executing state (blocked, dead) similarly deschedules the task.


### Spawning tasks

A call to `std::task::spawn`, passing a 0-argument function as its single
argument, causes the runtime to construct a new task executing the passed
function. The passed function is referred to as the _entry function_ for
the spawned task, and any captured environment is carries is moved from the
spawning task to the spawned task before the spawned task begins execution.

The result of a `spawn` call is a `std::task::task` value.

An example of a `spawn` call:

~~~~
import std::task::*;
import std::comm::*;

fn helper(c: chan<u8>) {
    // do some work.
    let result = ...;
    send(c, result);
}

let p: port<u8>;

spawn(bind helper(chan(p)));
// let task run, do other things.
// ...
let result = recv(p);
~~~~


### Sending values into channels

Sending a value into a channel is done by a library call to `std::comm::send`,
which takes a channel and a value to send, and moves the value into the
channel's outgoing buffer.

An example of a send:

~~~~
import std::comm::*;
let c: chan<str> = ...;
send(c, "hello, world");
~~~~


### Receiving values from ports

Receiving a value is done by a call to the `recv` method, on a value of type
`std::comm::port`. This call causes the receiving task to enter the *blocked
reading* state until a task is sending a value to the port, at which point the
runtime pseudo-randomly selects a sending task and moves a value from the head
of one of the task queues to the call's return value, and un-blocks the
receiving task. See [communication system](#communication-system).

An example of a *receive*:

~~~~~~~~
import std::comm::*;
let p: port<str> = ...;
let s: str = recv(p);
~~~~~~~~


# Runtime services, linkage and debugging

# Appendix: Rationales and design tradeoffs

*TODO*.

# Appendix: Influences and further references

## Influences


>  The essential problem that must be solved in making a fault-tolerant
>  software system is therefore that of fault-isolation. Different programmers
>  will write different modules, some modules will be correct, others will have
>  errors. We do not want the errors in one module to adversely affect the
>  behaviour of a module which does not have any errors.
>
>  &mdash; Joe Armstrong


>  In our approach, all data is private to some process, and processes can
>  only communicate through communications channels. *Security*, as used
>  in this paper, is the property which guarantees that processes in a system
>  cannot affect each other except by explicit communication.
>
>  When security is absent, nothing which can be proven about a single module
>  in isolation can be guaranteed to hold when that module is embedded in a
>  system [...]
>
>  &mdash;  Robert Strom and Shaula Yemini


>  Concurrent and applicative programming complement each other. The
>  ability to send messages on channels provides I/O without side effects,
>  while the avoidance of shared data helps keep concurrent processes from
>  colliding.
>
>  &mdash; Rob Pike


Rust is not a particularly original language. It may however appear unusual
by contemporary standards, as its design elements are drawn from a number of
"historical" languages that have, with a few exceptions, fallen out of
favour. Five prominent lineages contribute the most, though their influences
have come and gone during the course of Rust's development:

* The NIL (1981) and Hermes (1990) family. These languages were developed by
  Robert Strom, Shaula Yemini, David Bacon and others in their group at IBM
  Watson Research Center (Yorktown Heights, NY, USA).

* The Erlang (1987) language, developed by Joe Armstrong, Robert Virding, Claes
  Wikstr&ouml;m, Mike Williams and others in their group at the Ericsson Computer
  Science Laboratory (&Auml;lvsj&ouml;, Stockholm, Sweden) .

* The Sather (1990) language, developed by Stephen Omohundro, Chu-Cheow Lim,
  Heinz Schmidt and others in their group at The International Computer
  Science Institute of the University of California, Berkeley (Berkeley, CA,
  USA).

* The Newsqueak (1988), Alef (1995), and Limbo (1996) family. These
  languages were developed by Rob Pike, Phil Winterbottom, Sean Dorward and
  others in their group at Bell labs Computing Sciences Reserch Center
  (Murray Hill, NJ, USA).

* The Napier (1985) and Napier88 (1988) family. These languages were
  developed by Malcolm Atkinson, Ron Morrison and others in their group at
  the University of St. Andrews (St. Andrews, Fife, UK).

Additional specific influences can be seen from the following languages:

* The stack-growth implementation of Go.
* The structural algebraic types and compilation manager of SML.
* The attribute and assembly systems of C#.
* The deterministic destructor system of C++.
* The typeclass system of Haskell.
* The lexical identifier rule of Python.
* The block syntax of Ruby.