path: root/doc/go_spec.html

<!-- title The Go Programming Language Specification -->
<!-- subtitle Version of June 1, 2010 -->

<!--
TODO
[ ] need language about function/method calls and parameter passing rules
[ ] last paragraph of #Assignments (constant promotion) should be elsewhere
    and mention assignment to empty interface.
[ ] need to say something about "scope" of selectors?
[ ] clarify what a field name is in struct declarations
    (struct{T} vs struct {T T} vs struct {t T})
[ ] need explicit language about the result type of operations
[ ] may want to have some examples for the types of shift operations
[ ] should string(1<<s) and float(1<<s) be valid?
[ ] should probably write something about evaluation order of statements even
	though obvious
[ ] specify iteration direction for range clause
[ ] review language on implicit dereferencing
[ ] clarify what it means for two functions to be "the same" when comparing them
-->


<h2 id="Introduction">Introduction</h2>

<p>
This is a reference manual for the Go programming language. For
more information and other documents, see <a href="http://golang.org/">http://golang.org</a>.
</p>

<p>
Go is a general-purpose language designed with systems programming
in mind. It is strongly typed and garbage-collected and has explicit
support for concurrent programming.  Programs are constructed from
<i>packages</i>, whose properties allow efficient management of
dependencies. The existing implementations use a traditional
compile/link model to generate executable binaries.
</p>

<p>
The grammar is compact and regular, allowing for easy analysis by
automatic tools such as integrated development environments.
</p>

<h2 id="Notation">Notation</h2>
<p>
The syntax is specified using Extended Backus-Naur Form (EBNF):
</p>

<pre class="grammar">
Production  = production_name "=" Expression "." .
Expression  = Alternative { "|" Alternative } .
Alternative = Term { Term } .
Term        = production_name | token [ "..." token ] | Group | Option | Repetition .
Group       = "(" Expression ")" .
Option      = "[" Expression "]" .
Repetition  = "{" Expression "}" .
</pre>

<p>
Productions are expressions constructed from terms and the following
operators, in increasing precedence:
</p>
<pre class="grammar">
|   alternation
()  grouping
[]  option (0 or 1 times)
{}  repetition (0 to n times)
</pre>

<p>
Lower-case production names are used to identify lexical tokens.
Non-terminals are in CamelCase. Lexical symbols are enclosed in
double quotes <code>""</code> or back quotes <code>``</code>.
</p>

<p>
The form <code>a ... b</code> represents the set of characters from
<code>a</code> through <code>b</code> as alternatives.
</p>

<h2 id="Source_code_representation">Source code representation</h2>

<p>
Source code is Unicode text encoded in
<a href="http://en.wikipedia.org/wiki/UTF-8">UTF-8</a>. The text is not
canonicalized, so a single accented code point is distinct from the
same character constructed from combining an accent and a letter;
those are treated as two code points.  For simplicity, this document
will use the term <i>character</i> to refer to a Unicode code point.
</p>
<p>
Each code point is distinct; for instance, upper and lower case letters
are different characters.
</p>
<p>
Implementation restriction: For compatibility with other tools, a
compiler may disallow the NUL character (U+0000) in the source text.
</p>

<h3 id="Characters">Characters</h3>

<p>
The following terms are used to denote specific Unicode character classes:
</p>
<pre class="ebnf">
unicode_char   = /* an arbitrary Unicode code point */ .
unicode_letter = /* a Unicode code point classified as "Letter" */ .
unicode_digit  = /* a Unicode code point classified as "Digit" */ .
</pre>

<p>
In <a href="http://www.unicode.org/versions/Unicode5.2.0/">The Unicode Standard 5.2</a>,
Section 4.5 General Category-Normative
defines a set of character categories.  Go treats
those characters in category Lu, Ll, Lt, Lm, or Lo as Unicode letters,
and those in category Nd as Unicode digits.
</p>

<h3 id="Letters_and_digits">Letters and digits</h3>

<p>
The underscore character <code>_</code> (U+005F) is considered a letter.
</p>
<pre class="ebnf">
letter        = unicode_letter | "_" .
decimal_digit = "0" ... "9" .
octal_digit   = "0" ... "7" .
hex_digit     = "0" ... "9" | "A" ... "F" | "a" ... "f" .
</pre>

<h2 id="Lexical_elements">Lexical elements</h2>

<h3 id="Comments">Comments</h3>

<p>
There are two forms of comments:
</p>

<ol>
<li>
<i>Line comments</i> start with the character sequence <code>//</code>
and continue through the next newline. A line comment acts like a newline.
</li>
<li>
<i>General comments</i> start with the character sequence <code>/*</code>
and continue through the character sequence <code>*/</code>. A general
comment that spans multiple lines acts like a newline, otherwise it acts
like a space.
</li>
</ol>

<p>
Comments do not nest.
</p>


<h3 id="Tokens">Tokens</h3>

<p>
Tokens form the vocabulary of the Go language.
There are four classes: <i>identifiers</i>, <i>keywords</i>, <i>operators
and delimiters</i>, and <i>literals</i>.  <i>White space</i>, formed from
spaces (U+0020), horizontal tabs (U+0009),
carriage returns (U+000D), and newlines (U+000A),
is ignored except as it separates tokens
that would otherwise combine into a single token. Also, a newline
may trigger the insertion of a <a href="#Semicolons">semicolon</a>.
While breaking the input into tokens,
the next token is the longest sequence of characters that form a
valid token.
</p>

<h3 id="Semicolons">Semicolons</h3>

<p>
The formal grammar uses semicolons <code>";"</code> as terminators in
a number of productions. Go programs may omit most of these semicolons
using the following two rules:
</p>

<ol>
<li>
<p>
When the input is broken into tokens, a semicolon is automatically inserted
into the token stream at the end of a non-blank line if the line's final
token is
</p>
<ul>
	<li>an
	    <a href="#Identifiers">identifier</a>
	</li>
	
	<li>an
	    <a href="#Integer_literals">integer</a>,
	    <a href="#Floating-point_literals">floating-point</a>,
	    <a href="#Imaginary_literals">imaginary</a>,
	    <a href="#Character_literals">character</a>, or
	    <a href="#String_literals">string</a> literal
	</li>
	
	<li>one of the <a href="#Keywords">keywords</a>
	    <code>break</code>,
	    <code>continue</code>,
	    <code>fallthrough</code>, or
	    <code>return</code>
	</li>
	
	<li>one of the <a href="#Operators_and_Delimiters">operators and delimiters</a>
	    <code>++</code>,
	    <code>--</code>,
	    <code>)</code>,
	    <code>]</code>, or
	    <code>}</code>
	</li>
</ul>
</li>

<li>
To allow complex statements to occupy a single line, a semicolon
may be omitted before a closing <code>")"</code> or <code>"}"</code>.
</li>
</ol>

<p>
To reflect idiomatic use, code examples in this document elide semicolons
using these rules.
</p>


<h3 id="Identifiers">Identifiers</h3>

<p>
Identifiers name program entities such as variables and types.
An identifier is a sequence of one or more letters and digits.
The first character in an identifier must be a letter.
</p>
<pre class="ebnf">
identifier = letter { letter | unicode_digit } .
</pre>
<pre>
a
_x9
ThisVariableIsExported
αβ
</pre>

<p>
Some identifiers are <a href="#Predeclared_identifiers">predeclared</a>.
</p>


<h3 id="Keywords">Keywords</h3>

<p>
The following keywords are reserved and may not be used as identifiers.
</p>
<pre class="grammar">
break        default      func         interface    select
case         defer        go           map          struct
chan         else         goto         package      switch
const        fallthrough  if           range        type
continue     for          import       return       var
</pre>

<h3 id="Operators_and_Delimiters">Operators and Delimiters</h3>

<p>
The following character sequences represent <a href="#Operators">operators</a>, delimiters, and other special tokens:
</p>
<pre class="grammar">
+    &amp;     +=    &amp;=     &amp;&amp;    ==    !=    (    )
-    |     -=    |=     ||    &lt;     &lt;=    [    ]
*    ^     *=    ^=     &lt;-    &gt;     &gt;=    {    }
/    &lt;&lt;    /=    &lt;&lt;=    ++    =     :=    ,    ;
%    &gt;&gt;    %=    &gt;&gt;=    --    !     ...   .    :
     &amp;^          &amp;^=
</pre>

<h3 id="Integer_literals">Integer literals</h3>

<p>
An integer literal is a sequence of digits representing an
<a href="#Constants">integer constant</a>.
An optional prefix sets a non-decimal base: <code>0</code> for octal, <code>0x</code> or
<code>0X</code> for hexadecimal.  In hexadecimal literals, letters
<code>a-f</code> and <code>A-F</code> represent values 10 through 15.
</p>
<pre class="ebnf">
int_lit     = decimal_lit | octal_lit | hex_lit .
decimal_lit = ( "1" ... "9" ) { decimal_digit } .
octal_lit   = "0" { octal_digit } .
hex_lit     = "0" ( "x" | "X" ) hex_digit { hex_digit } .
</pre>

<pre>
42
0600
0xBadFace
170141183460469231731687303715884105727
</pre>

<h3 id="Floating-point_literals">Floating-point literals</h3>
<p>
A floating-point literal is a decimal representation of a
<a href="#Constants">floating-point constant</a>.
It has an integer part, a decimal point, a fractional part,
and an exponent part.  The integer and fractional part comprise
decimal digits; the exponent part is an <code>e</code> or <code>E</code>
followed by an optionally signed decimal exponent.  One of the
integer part or the fractional part may be elided; one of the decimal
point or the exponent may be elided.
</p>
<pre class="ebnf">
float_lit = decimals "." [ decimals ] [ exponent ] |
            decimals exponent |
            "." decimals [ exponent ] .
decimals  = decimal_digit { decimal_digit } .
exponent  = ( "e" | "E" ) [ "+" | "-" ] decimals .
</pre>

<pre>
0.
72.40
072.40  // == 72.40
2.71828
1.e+0
6.67428e-11
1E6
.25
.12345E+5
</pre>

<h3 id="Imaginary_literals">Imaginary literals</h3>
<p>
An imaginary literal is a decimal representation of the imaginary part of a
<a href="#Constants">complex constant</a>.
It consists of a
<a href="#Floating-point_literals">floating-point literal</a>
or decimal integer followed
by the lower-case letter <code>i</code>.
</p>
<pre class="ebnf">
imaginary_lit = (decimals | float_lit) "i" .
</pre>

<pre>
0i
011i  // == 11i
0.i
2.71828i
1.e+0i
6.67428e-11i
1E6i
.25i
.12345E+5i
</pre>


<h3 id="Character_literals">Character literals</h3>

<p>
A character literal represents an <a href="#Constants">integer constant</a>,
typically a Unicode code point, as one or more characters enclosed in single
quotes.  Within the quotes, any character may appear except single
quote and newline. A single quoted character represents itself,
while multi-character sequences beginning with a backslash encode
values in various formats.
</p>
<p>
The simplest form represents the single character within the quotes;
since Go source text is Unicode characters encoded in UTF-8, multiple
UTF-8-encoded bytes may represent a single integer value.  For
instance, the literal <code>'a'</code> holds a single byte representing
a literal <code>a</code>, Unicode U+0061, value <code>0x61</code>, while
<code>'ä'</code> holds two bytes (<code>0xc3</code> <code>0xa4</code>) representing
a literal <code>a</code>-dieresis, U+00E4, value <code>0xe4</code>.
</p>
<p>
Several backslash escapes allow arbitrary values to be represented
as ASCII text.  There are four ways to represent the integer value
as a numeric constant: <code>\x</code> followed by exactly two hexadecimal
digits; <code>\u</code> followed by exactly four hexadecimal digits;
<code>\U</code> followed by exactly eight hexadecimal digits, and a
plain backslash <code>\</code> followed by exactly three octal digits.
In each case the value of the literal is the value represented by
the digits in the corresponding base.
</p>
<p>
Although these representations all result in an integer, they have
different valid ranges.  Octal escapes must represent a value between
0 and 255 inclusive.  Hexadecimal escapes satisfy this condition
by construction. The escapes <code>\u</code> and <code>\U</code>
represent Unicode code points so within them some values are illegal,
in particular those above <code>0x10FFFF</code> and surrogate halves.
</p>
<p>
After a backslash, certain single-character escapes represent special values:
</p>
<pre class="grammar">
\a   U+0007 alert or bell
\b   U+0008 backspace
\f   U+000C form feed
\n   U+000A line feed or newline
\r   U+000D carriage return
\t   U+0009 horizontal tab
\v   U+000b vertical tab
\\   U+005c backslash
\'   U+0027 single quote  (valid escape only within character literals)
\"   U+0022 double quote  (valid escape only within string literals)
</pre>
<p>
All other sequences starting with a backslash are illegal inside character literals.
</p>
<pre class="ebnf">
char_lit         = "'" ( unicode_value | byte_value ) "'" .
unicode_value    = unicode_char | little_u_value | big_u_value | escaped_char .
byte_value       = octal_byte_value | hex_byte_value .
octal_byte_value = `\` octal_digit octal_digit octal_digit .
hex_byte_value   = `\` "x" hex_digit hex_digit .
little_u_value   = `\` "u" hex_digit hex_digit hex_digit hex_digit .
big_u_value      = `\` "U" hex_digit hex_digit hex_digit hex_digit
                           hex_digit hex_digit hex_digit hex_digit .
escaped_char     = `\` ( "a" | "b" | "f" | "n" | "r" | "t" | "v" | `\` | "'" | `"` ) .
</pre>

<pre>
'a'
'ä'
'本'
'\t'
'\000'
'\007'
'\377'
'\x07'
'\xff'
'\u12e4'
'\U00101234'
</pre>


<h3 id="String_literals">String literals</h3>

<p>
A string literal represents a <a href="#Constants">string constant</a>
obtained from concatenating a sequence of characters. There are two forms:
raw string literals and interpreted string literals.
</p>
<p>
Raw string literals are character sequences between back quotes
<code>``</code>.  Within the quotes, any character is legal except
back quote. The value of a raw string literal is the
string composed of the uninterpreted characters between the quotes;
in particular, backslashes have no special meaning and the string may
span multiple lines.
</p>
<p>
Interpreted string literals are character sequences between double
quotes <code>&quot;&quot;</code>. The text between the quotes,
which may not span multiple lines, forms the
value of the literal, with backslash escapes interpreted as they
are in character literals (except that <code>\'</code> is illegal and
<code>\"</code> is legal).  The three-digit octal (<code>\</code><i>nnn</i>)
and two-digit hexadecimal (<code>\x</code><i>nn</i>) escapes represent individual
<i>bytes</i> of the resulting string; all other escapes represent
the (possibly multi-byte) UTF-8 encoding of individual <i>characters</i>.
Thus inside a string literal <code>\377</code> and <code>\xFF</code> represent
a single byte of value <code>0xFF</code>=255, while <code>ÿ</code>,
<code>\u00FF</code>, <code>\U000000FF</code> and <code>\xc3\xbf</code> represent
the two bytes <code>0xc3</code> <code>0xbf</code> of the UTF-8 encoding of character
U+00FF.
</p>

<pre class="ebnf">
string_lit             = raw_string_lit | interpreted_string_lit .
raw_string_lit         = "`" { unicode_char } "`" .
interpreted_string_lit = `"` { unicode_value | byte_value } `"` .
</pre>

<pre>
`abc`  // same as "abc"
`\n
\n`    // same as "\\n\n\\n"
"\n"
""
"Hello, world!\n"
"日本語"
"\u65e5本\U00008a9e"
"\xff\u00FF"
</pre>

<p>
These examples all represent the same string:
</p>

<pre>
"日本語"                                 // UTF-8 input text
`日本語`                                 // UTF-8 input text as a raw literal
"\u65e5\u672c\u8a9e"                    // The explicit Unicode code points
"\U000065e5\U0000672c\U00008a9e"        // The explicit Unicode code points
"\xe6\x97\xa5\xe6\x9c\xac\xe8\xaa\x9e"  // The explicit UTF-8 bytes
</pre>

<p>
If the source code represents a character as two code points, such as
a combining form involving an accent and a letter, the result will be
an error if placed in a character literal (it is not a single code
point), and will appear as two code points if placed in a string
literal.
</p>


<h2 id="Constants">Constants</h2>

<p>There are <i>boolean constants</i>, <i>integer constants</i>,
<i>floating-point constants</i>, <i>complex constants</i>,
and <i>string constants</i>. Integer, floating-point,
and complex constants are
collectively called <i>numeric constants</i>.
</p>

<p>
A constant value is represented by an
<a href="#Integer_literals">integer</a>,
<a href="#Floating-point_literals">floating-point</a>,
<a href="#Imaginary_literals">imaginary</a>,
<a href="#Character_literals">character</a>, or
<a href="#String_literals">string</a> literal,
an identifier denoting a constant,
a <a href="#Constant_expressions">constant expression</a>, or
the result value of some built-in functions such as <code>unsafe.Sizeof</code>
and <code>cap</code> or <code>len</code> applied to an array,
<code>len</code> applied to a string constant,
<code>real</code> and <code>imag</code> applied to a complex constant
and <code>cmplx</code> applied to numeric constants.
The boolean truth values are represented by the predeclared constants
<code>true</code> and <code>false</code>. The predeclared identifier
<a href="#Iota">iota</a> denotes an integer constant.
</p>

<p>
In general, complex constants are a form of
<a href="#Constant_expressions">constant expression</a>
and are discussed in that section.
</p>

<p>
Numeric constants represent values of arbitrary precision and do not overflow.
</p>

<p>
Constants may be <a href="#Types">typed</a> or untyped.
Literal constants, <code>true</code>, <code>false</code>, <code>iota</code>,
and certain <a href="#Constant_expressions">constant expressions</a>
containing only untyped constant operands are untyped.
</p>

<p>
A constant may be given a type explicitly by a <a href="#Constant_declarations">constant declaration</a>
or <a href="#Conversions">conversion</a>, or implicitly when used in a
<a href="#Variable_declarations">variable declaration</a> or an
<a href="#Assignments">assignment</a> or as an
operand in an <a href="#Expressions">expression</a>.
It is an error if the constant value
cannot be accurately represented as a value of the respective type.
For instance, <code>3.0</code> can be given any integer or any
floating-point type, while <code>2147483648.0</code> (equal to <code>1&lt;&lt;31</code>)
can be given the types <code>float32</code>, <code>float64</code>, or <code>uint32</code> but
not <code>int32</code> or <code>string</code>.
</p>

<p>
There are no constants denoting the IEEE-754 infinity and not-a-number values,
but the <a href="/pkg/math/"><code>math</code> package</a>'s
<a href="/pkg/math/#Inf">Inf</a>,
<a href="/pkg/math/#NaN">NaN</a>,
<a href="/pkg/math/#IsInf">IsInf</a>, and
<a href="/pkg/math/#IsNaN">IsNaN</a>
functions return and test for those values at run time.
</p>

<p>
Implementation restriction: A compiler may implement numeric constants by choosing
an internal representation with at least twice as many bits as any machine type;
for floating-point values, both the mantissa and exponent must be twice as large.
</p>


<h2 id="Types">Types</h2>

<p>
A type determines the set of values and operations specific to values of that
type.  A type may be specified by a (possibly qualified) <i>type name</i>
(§<a href="#Qualified_identifier">Qualified identifier</a>, §<a href="#Type_declarations">Type declarations</a>) or a <i>type literal</i>,
which composes a new type from previously declared types.
</p>

<pre class="ebnf">
Type      = TypeName | TypeLit | "(" Type ")" .
TypeName  = QualifiedIdent.
TypeLit   = ArrayType | StructType | PointerType | FunctionType | InterfaceType |
	    SliceType | MapType | ChannelType .
</pre>

<p>
Named instances of the boolean, numeric, and string types are
<a href="#Predeclared_identifiers">predeclared</a>.
<i>Composite types</i>&mdash;array, struct, pointer, function,
interface, slice, map, and channel types&mdash;may be constructed using
type literals.
</p>

<p>
A type may have a <i>method set</i> associated with it
(§<a href="#Interface_types">Interface types</a>, §<a href="#Method_declarations">Method declarations</a>).
The method set of an <a href="#Interface_types">interface type</a> is its interface.
The method set of any other named type <code>T</code>
consists of all methods with receiver type <code>T</code>.
The method set of the corresponding pointer type <code>*T</code>
is the set of all methods with receiver <code>*T</code> or <code>T</code>
(that is, it also contains the method set of <code>T</code>).
Any other type has an empty method set.
In a method set, each method must have a unique name.
</p>
<p>
The <i>static type</i> (or just <i>type</i>) of a variable is the
type defined by its declaration.  Variables of interface type
also have a distinct <i>dynamic type</i>, which
is the actual type of the value stored in the variable at run-time.
The dynamic type may vary during execution but is always assignment compatible
to the static type of the interface variable.  For non-interface
types, the dynamic type is always the static type.
</p>


<h3 id="Boolean_types">Boolean types</h3>

A <i>boolean type</i> represents the set of Boolean truth values
denoted by the predeclared constants <code>true</code>
and <code>false</code>. The predeclared boolean type is <code>bool</code>.


<h3 id="Numeric_types">Numeric types</h3>

<p>
A <i>numeric type</i> represents sets of integer or floating-point values.
The predeclared architecture-independent numeric types are:
</p>

<pre class="grammar">
uint8       the set of all unsigned  8-bit integers (0 to 255)
uint16      the set of all unsigned 16-bit integers (0 to 65535)
uint32      the set of all unsigned 32-bit integers (0 to 4294967295)
uint64      the set of all unsigned 64-bit integers (0 to 18446744073709551615)

int8        the set of all signed  8-bit integers (-128 to 127)
int16       the set of all signed 16-bit integers (-32768 to 32767)
int32       the set of all signed 32-bit integers (-2147483648 to 2147483647)
int64       the set of all signed 64-bit integers (-9223372036854775808 to 9223372036854775807)

float32     the set of all IEEE-754 32-bit floating-point numbers
float64     the set of all IEEE-754 64-bit floating-point numbers

complex64   the set of all complex numbers with float32 real and imaginary parts
complex128  the set of all complex numbers with float64 real and imaginary parts

byte        familiar alias for uint8
</pre>

<p>
The value of an <i>n</i>-bit integer is <i>n</i> bits wide and represented using
<a href="http://en.wikipedia.org/wiki/Two's_complement">two's complement arithmetic</a>.
</p>

<p>
There is also a set of predeclared numeric types with implementation-specific sizes:
</p>

<pre class="grammar">
uint     either 32 or 64 bits
int      either 32 or 64 bits
float    either 32 or 64 bits
complex  real and imaginary parts have type float
uintptr  an unsigned integer large enough to store the uninterpreted bits of a pointer value
</pre>

<p>
To avoid portability issues all numeric types are distinct except
<code>byte</code>, which is an alias for <code>uint8</code>.
Conversions
are required when incompatible numeric types are mixed in an expression
or assignment. For instance, <code>int32</code> and <code>int</code>
are not the same type even though they may have the same size on a
particular architecture.


<h3 id="String_types">String types</h3>

<p>
A <i>string type</i> represents the set of string values.
Strings behave like arrays of bytes but are immutable: once created,
it is impossible to change the contents of a string.
The predeclared string type is <code>string</code>.

<p>
The elements of strings have type <code>byte</code> and may be
accessed using the usual <a href="#Indexes">indexing operations</a>.  It is
illegal to take the address of such an element; if
<code>s[i]</code> is the <i>i</i>th byte of a
string, <code>&amp;s[i]</code> is invalid.  The length of string
<code>s</code> can be discovered using the built-in function
<code>len</code>. The length is a compile-time constant if <code>s</code>
is a string literal.
</p>


<h3 id="Array_types">Array types</h3>

<p>
An array is a numbered sequence of elements of a single
type, called the element type.
The number of elements is called the length and is never
negative.
</p>

<pre class="ebnf">
ArrayType   = "[" ArrayLength "]" ElementType .
ArrayLength = Expression .
ElementType = Type .
</pre>

<p>
The length is part of the array's type and must be a
<a href="#Constant_expressions">constant expression</a> that evaluates to a non-negative
integer value.  The length of array <code>a</code> can be discovered
using the built-in function <code>len(a)</code>, which is a
compile-time constant.  The elements can be indexed by integer
indices 0 through the <code>len(a)-1</code> (§<a href="#Indexes">Indexes</a>).
Array types are always one-dimensional but may be composed to form
multi-dimensional types.
</p>

<pre>
[32]byte
[2*N] struct { x, y int32 }
[1000]*float64
[3][5]int
[2][2][2]float64  // same as [2]([2]([2]float64))
</pre>

<h3 id="Slice_types">Slice types</h3>

<p>
A slice is a reference to a contiguous segment of an array and
contains a numbered sequence of elements from that array.  A slice
type denotes the set of all slices of arrays of its element type.
A slice value may be <code>nil</code>.
</p>

<pre class="ebnf">
SliceType = "[" "]" ElementType .
</pre>

<p>
Like arrays, slices are indexable and have a length.  The length of a
slice <code>s</code> can be discovered by the built-in function
<code>len(s)</code>; unlike with arrays it may change during
execution.  The elements can be addressed by integer indices 0
through <code>len(s)-1</code> (§<a href="#Indexes">Indexes</a>).  The slice index of a
given element may be less than the index of the same element in the
underlying array.
</p>
<p>
A slice, once initialized, is always associated with an underlying
array that holds its elements.  A slice therefore shares storage
with its array and with other slices of the same array; by contrast,
distinct arrays always represent distinct storage.
</p>
<p>
The array underlying a slice may extend past the end of the slice.
The <i>capacity</i> is a measure of that extent: it is the sum of
the length of the slice and the length of the array beyond the slice;
a slice of length up to that capacity can be created by `slicing' a new
one from the original slice (§<a href="#Slices">Slices</a>).
The capacity of a slice <code>a</code> can be discovered using the
built-in function <code>cap(a)</code> and the relationship between
<code>len()</code> and <code>cap()</code> is:
</p>

<pre>
0 <= len(a) <= cap(a)
</pre>

<p>
The value of an uninitialized slice is <code>nil</code>.
The length and capacity of a <code>nil</code> slice
are 0. A new, initialized slice value for a given element type <code>T</code> is
made using the built-in function <code>make</code>, which takes a slice type
and parameters specifying the length and optionally the capacity:
</p>

<pre>
make([]T, length)
make([]T, length, capacity)
</pre>

<p>
The <code>make()</code> call allocates a new, hidden array to which the returned
slice value refers. That is, executing
</p>

<pre>
make([]T, length, capacity)
</pre>

<p>
produces the same slice as allocating an array and slicing it, so these two examples
result in the same slice:
</p>

<pre>
make([]int, 50, 100)
new([100]int)[0:50]
</pre>

<p>
Like arrays, slices are always one-dimensional but may be composed to construct
higher-dimensional objects.
With arrays of arrays, the inner arrays are, by construction, always the same length;
however with slices of slices (or arrays of slices), the lengths may vary dynamically.
Moreover, the inner slices must be allocated individually (with <code>make</code>).
</p>

<h3 id="Struct_types">Struct types</h3>

<p>
A struct is a sequence of named elements, called fields, each of which has a
name and a type. Field names may be specified explicitly (IdentifierList) or
implicitly (AnonymousField).
Within a struct, non-<a href="#Blank_identifier">blank</a> field names must
be unique.
</p>

<pre class="ebnf">
StructType     = "struct" "{" { FieldDecl ";" } "}" .
FieldDecl      = (IdentifierList Type | AnonymousField) [ Tag ] .
AnonymousField = [ "*" ] TypeName .
Tag            = string_lit .
</pre>

<pre>
// An empty struct.
struct {}

// A struct with 6 fields.
struct {
	x, y int
	u float
	_ float  // padding
	A *[]int
	F func()
}
</pre>

<p>
A field declared with a type but no explicit field name is an <i>anonymous field</i>.
Such a field type must be specified as
a type name <code>T</code> or as a pointer to a type name <code>*T</code>,
and <code>T</code> itself may not be
a pointer type. The unqualified type name acts as the field name.
</p>

<pre>
// A struct with four anonymous fields of type T1, *T2, P.T3 and *P.T4
struct {
	T1        // field name is T1
	*T2       // field name is T2
	P.T3      // field name is T3
	*P.T4     // field name is T4
	x, y int  // field names are x and y
}
</pre>

<p>
The following declaration is illegal because field names must be unique
in a struct type:
</p>

<pre>
struct {
	T         // conflicts with anonymous field *T and *P.T
	*T        // conflicts with anonymous field T and *P.T
	*P.T      // conflicts with anonymous field T and *T
}
</pre>

<p>
Fields and methods (§<a href="#Method_declarations">Method declarations</a>) of an anonymous field are
promoted to be ordinary fields and methods of the struct (§<a href="#Selectors">Selectors</a>).
The following rules apply for a struct type named <code>S</code> and
a type named <code>T</code>:
</p>
<ul>
	<li>If <code>S</code> contains an anonymous field <code>T</code>, the
	    method set of <code>S</code> includes the method set of <code>T</code>.
	</li>

	<li>If <code>S</code> contains an anonymous field <code>*T</code>, the
	    method set of <code>S</code> includes the method set of <code>*T</code>
	    (which itself includes the method set of <code>T</code>).
	</li>

	<li>If <code>S</code> contains an anonymous field <code>T</code> or
	    <code>*T</code>, the method set of <code>*S</code> includes the
	    method set of <code>*T</code> (which itself includes the method
	    set of <code>T</code>).
	</li>
</ul>
<p>
A field declaration may be followed by an optional string literal <i>tag</i>,
which becomes an attribute for all the fields in the corresponding
field declaration. The tags are made
visible through a <a href="#Package_unsafe">reflection interface</a>
but are otherwise ignored.
</p>

<pre>
// A struct corresponding to the TimeStamp protocol buffer.
// The tag strings define the protocol buffer field numbers.
struct {
	microsec  uint64 "field 1"
	serverIP6 uint64 "field 2"
	process   string "field 3"
}
</pre>

<h3 id="Pointer_types">Pointer types</h3>

<p>
A pointer type denotes the set of all pointers to variables of a given
type, called the <i>base type</i> of the pointer.
A pointer value may be <code>nil</code>.
</p>

<pre class="ebnf">
PointerType = "*" BaseType .
BaseType = Type .
</pre>

<pre>
*int
*map[string] *chan int
</pre>

<h3 id="Function_types">Function types</h3>

<p>
A function type denotes the set of all functions with the same parameter
and result types.
A function value may be <code>nil</code>.
</p>

<pre class="ebnf">
FunctionType   = "func" Signature .
Signature      = Parameters [ Result ] .
Result         = Parameters | Type .
Parameters     = "(" [ ParameterList [ "," ] ] ")" .
ParameterList  = ParameterDecl { "," ParameterDecl } .
ParameterDecl  = [ IdentifierList ] ( Type | "..." [ Type ] ) .
</pre>

<p>
Within a list of parameters or results, the names (IdentifierList)
must either all be present or all be absent. If present, each name
stands for one item (parameter or result) of the specified type; if absent, each
type stands for one item of that type.  Parameter and result
lists are always parenthesized except that if there is exactly
one unnamed result it may written as an unparenthesized type.
</p>
<p>
For the last parameter only, instead of a type one may write
<code>...</code> or <code>...  T</code> to indicate that the function
may be invoked with zero or more additional arguments.  If the type
<code>T</code> is present in the parameter declaration, the additional
arguments must all be
<a href="#Assignment_compatibility">assignment compatible</a>
with type <code>T</code>; otherwise they may be of any type.
</p>

<pre>
func()
func(x int)
func() int
func(string, float, ...)
func(prefix string, values ... int)
func(a, b int, z float) bool
func(a, b int, z float) (bool)
func(a, b int, z float, opt ...) (success bool)
func(int, int, float) (float, *[]int)
func(n int) func(p *T)
</pre>


<h3 id="Interface_types">Interface types</h3>

<p>
An interface type specifies a <a href="#Types">method set</a> called its <i>interface</i>.
A variable of interface type can store a value of any type with a method set
that is any superset of the interface. Such a type is said to
<i>implement the interface</i>. An interface value may be <code>nil</code>.
</p>

<pre class="ebnf">
InterfaceType      = "interface" "{" { MethodSpec ";" } "}" .
MethodSpec         = MethodName Signature | InterfaceTypeName .
MethodName         = identifier .
InterfaceTypeName  = TypeName .
</pre>

<p>
As with all method sets, in an interface type, each method must have a unique name.
</p>

<pre>
// A simple File interface
interface {
	Read(b Buffer) bool
	Write(b Buffer) bool
	Close()
}
</pre>

<p>
More than one type may implement an interface.
For instance, if two types <code>S1</code> and <code>S2</code>
have the method set
</p>

<pre>
func (p T) Read(b Buffer) bool { return ... }
func (p T) Write(b Buffer) bool { return ... }
func (p T) Close() { ... }
</pre>

<p>
(where <code>T</code> stands for either <code>S1</code> or <code>S2</code>)
then the <code>File</code> interface is implemented by both <code>S1</code> and
<code>S2</code>, regardless of what other methods
<code>S1</code> and <code>S2</code> may have or share.
</p>

<p>
A type implements any interface comprising any subset of its methods
and may therefore implement several distinct interfaces. For
instance, all types implement the <i>empty interface</i>:
</p>

<pre>
interface{}
</pre>

<p>
Similarly, consider this interface specification,
which appears within a <a href="#Type_declarations">type declaration</a>
to define an interface called <code>Lock</code>:
</p>

<pre>
type Lock interface {
	Lock()
	Unlock()
}
</pre>

<p>
If <code>S1</code> and <code>S2</code> also implement
</p>

<pre>
func (p T) Lock() { ... }
func (p T) Unlock() { ... }
</pre>

<p>
they implement the <code>Lock</code> interface as well
as the <code>File</code> interface.
</p>
<p>
An interface may contain an interface type name <code>T</code>
in place of a method specification.
The effect is equivalent to enumerating the methods of <code>T</code> explicitly
in the interface.
</p>

<pre>
type ReadWrite interface {
	Read(b Buffer) bool
	Write(b Buffer) bool
}

type File interface {
	ReadWrite  // same as enumerating the methods in ReadWrite
	Lock       // same as enumerating the methods in Lock
	Close()
}
</pre>

<h3 id="Map_types">Map types</h3>

<p>
A map is an unordered group of elements of one type, called the
element type, indexed by a set of unique <i>keys</i> of another type,
called the key type.
A map value may be <code>nil</code>.

</p>

<pre class="ebnf">
MapType     = "map" "[" KeyType "]" ElementType .
KeyType     = Type .
</pre>

<p>
The comparison operators <code>==</code> and <code>!=</code>
(§<a href="#Comparison_operators">Comparison operators</a>) must be fully defined
for operands of the key type; thus the key type must not be a struct, array or slice.
If the key type is an interface type, these
comparison operators must be defined for the dynamic key values;
failure will cause a <a href="#Run_time_panics">run-time panic</a>.

</p>

<pre>
map [string] int
map [*T] struct { x, y float }
map [string] interface {}
</pre>

<p>
The number of elements is called the length and is never negative.
The length of a map <code>m</code> can be discovered using the
built-in function <code>len(m)</code> and may change during execution.
Values may be added and removed
during execution using special forms of <a href="#Assignments">assignment</a>.
</p>
<p>
The value of an uninitialized map is <code>nil</code>.
A new, empty map value is made using the built-in
function <code>make</code>, which takes the map type and an optional
capacity hint as arguments:
</p>

<pre>
make(map[string] int)
make(map[string] int, 100)
</pre>

<p>
The initial capacity does not bound its size:
maps grow to accommodate the number of items
stored in them.
</p>

<h3 id="Channel_types">Channel types</h3>

<p>
A channel provides a mechanism for two concurrently executing functions
to synchronize execution and communicate by passing a value of a
specified element type.
A value of channel type may be <code>nil</code>.
</p>

<pre class="ebnf">
ChannelType = ( "chan" [ "&lt;-" ] | "&lt;-" "chan" ) ElementType .
</pre>

<p>
The <code>&lt;-</code> operator specifies the channel <i>direction</i>,
<i>send</i> or <i>receive</i>. If no direction is given, the channel is
<i>bi-directional</i>.
A channel may be constrained only to send or only to receive by
<a href="#Conversions">conversion</a> or <a href="#Assignments">assignment</a>.
</p>

<pre>
chan T         // can be used to send and receive values of type T
chan&lt;- float   // can only be used to send floats
&lt;-chan int     // can only be used to receive ints
</pre>

<p>
The <code>&lt;-</code> operator associates with the leftmost <code>chan</code>
possible:
</p>

<pre>
chan&lt;- chan int     // same as chan&lt;- (chan int)
chan&lt;- &lt;-chan int   // same as chan&lt;- (&lt;-chan int)
&lt;-chan &lt;-chan int   // same as &lt;-chan (&lt;-chan int)
chan (&lt;-chan int)
</pre>

<p>
The value of an uninitialized channel is <code>nil</code>. A new, initialized channel
value can be made using the built-in function
<a href="#Making_slices_maps_and_channels"><code>make</code></a>,
which takes the channel type and an optional capacity as arguments:
</p>

<pre>
make(chan int, 100)
</pre>

<p>
The capacity, in number of elements, sets the size of the buffer in the channel. If the
capacity is greater than zero, the channel is asynchronous: provided the
buffer is not full, sends can succeed without blocking. If the capacity is zero
or absent, the communication succeeds only when both a sender and receiver are ready.
</p>

<p>
A channel may be closed and tested for closure with the built-in functions
<a href="#Close_and_closed"><code>close</code> and <code>closed</code></a>.
</p>

<h2 id="Properties_of_types_and_values">Properties of types and values</h2>

<p>
Two types are either <i>identical</i> or <i>different</i>, and they are
either <i>compatible</i> or <i>incompatible</i>.
Identical types are always compatible, but compatible types need not be identical.
</p>

<h3 id="Type_identity_and_compatibility">Type identity and compatibility</h3>

<h4 id="Type_identity">Type identity</h4>

<p>
Two named types are identical if their type names originate in the same
type declaration (§<a href="#Declarations_and_scope">Declarations and scope</a>).
A named and an unnamed type are always different. Two unnamed types are identical
if the corresponding type literals are identical; that is if they have the same
literal structure and corresponding components have identical types. In detail:
</p>

<ul>
	<li>Two array types are identical if they have identical element types and
	    the same array length.</li>

	<li>Two slice types are identical if they have identical element types.</li>

	<li>Two struct types are identical if they have the same sequence of fields,
	    and if corresponding fields have the same names and identical types.
	    Two anonymous fields are considered to have the same name. Lower-case field
	    names from different packages are always different.</li>

	<li>Two pointer types are identical if they have identical base types.</li>

	<li>Two function types are identical if they have the same number of parameters
	    and result values and if corresponding parameter and result types are
	    identical. All "..." parameters without a specified type are defined to have
	    identical type.  All "..." parameters with specified identical type
	    <code>T</code> are defined to have identical type.
	    Parameter and result names are not required to match.</li>

	<li>Two interface types are identical if they have the same set of methods
	    with the same names and identical function types. Lower-case method names from
	    different packages are always different. The order of the methods is irrelevant.</li>

	<li>Two map types are identical if they have identical key and value types.</li>

	<li>Two channel types are identical if they have identical value types and
	    the same direction.</li>
</ul>

<h4 id="Type_compatibility">Type compatibility</h4>

<p>
Type compatibility is less stringent than type identity: All identical types are
compatible, but additionally a named and an unnamed type are compatible if the
respective type literals are identical.
</p>

<p>
Given the declarations
</p>

<pre>
type (
	T0 []string
	T1 []string
	T2 struct { a, b int }
	T3 struct { a, c int }
	T4 func(int, float) *T0
	T5 func(x int, y float) *[]string
)
</pre>

<p>
these types are identical:
</p>

<pre>
T0 and T0
[]int and []int
struct { a, b *T5 } and struct { a, b *T5 }
func(x int, y float) *[]string and func(int, float) (result *[]string)
</pre>

<p>
<code>T0</code> and <code>T1</code> are neither identical nor compatible
because they are named types with distinct declarations.
</p>

<p>
These types are compatible:
</p>

<pre>
T0 and T0
T0 and []string
T3 and struct { a int; c int }
T4 and func(x int, y float) (result *T0)
</pre>

<p>
<code>T2</code> and <code>struct { a, c int }</code> are incompatible because
they have different field names; <code>T4</code> and
<code>func(x int, y float) *[]string</code> are incompatible because the
respective type literals are different.
</p>

<h3 id="Assignment_compatibility">Assignment compatibility</h3>

<p>
A value <code>v</code> of static type <code>V</code> is <i>assignment compatible</i>
with a type <code>T</code> if one or more of the following conditions applies:
</p>

<ul>
<li>
<code>V</code> is compatible with <code>T</code>.
</li>
<li>
<code>T</code> is an interface type and
<code>V</code> <a href="#Interface_types">implements</a> <code>T</code>.
</li>
<li>
<code>V</code> is a bidirectional channel and <code>T</code> is a channel type
with identical element type and at least one of <code>V</code> or <code>T</code> is unnamed.
</li>
</ul>

<p>
If <code>T</code> is a struct type, either all fields of <code>T</code>
must be <a href="#Exported_identifiers">exported</a>, or the assignment must be in
the same package in which <code>T</code> is declared.
In other words, a struct value can be assigned to a struct variable only if
every field of the struct may be legally assigned individually by the program.
</p>

<p>
An untyped <a href="#Constants">constant</a> <code>v</code>
is assignment compatible with type <code>T</code> if <code>v</code>
can be represented accurately as a value of type <code>T</code>.
</p>

<p>
The predeclared identifier <code>nil</code> is assignment compatible with any
pointer, function, slice, map, channel, or interface type and
represents the <a href="#The_zero_value">zero value</a> for that type.
</p>

<p>
Any value may be assigned to the <a href="#Blank_identifier">blank identifier</a>.
</p>


<h2 id="Blocks">Blocks</h2>

<p>
A <i>block</i> is a sequence of declarations and statements within matching
brace brackets.
</p>

<pre class="ebnf">
Block = "{" { Statement ";" } "}" .
</pre>

<p>
In addition to explicit blocks in the source code, there are implicit blocks:
</p>

<ol>
	<li>The <i>universe block</i> encompasses all Go source text.</li>

	<li>Each <a href="#Packages">package</a> has a <i>package block</i> containing all
	    Go source text for that package.</li>

	<li>Each file has a <i>file block</i> containing all Go source text
	    in that file.</li>

	<li>Each <code>if</code>, <code>for</code>, and <code>switch</code>
	    statement is considered to be in its own implicit block.</li>

	<li>Each clause in a <code>switch</code> or <code>select</code> statement
	    acts as an implicit block.</li>
</ol>

<p>
Blocks nest and influence <a href="#Declarations_and_scope">scoping</a>.
</p>


<h2 id="Declarations_and_scope">Declarations and scope</h2>

<p>
A declaration binds a non-<a href="#Blank_identifier">blank</a>
identifier to a constant, type, variable, function, or package.
Every identifier in a program must be declared.
No identifier may be declared twice in the same block, and
no identifier may be declared in both the file and package block.
</p>

<pre class="ebnf">
Declaration   = ConstDecl | TypeDecl | VarDecl .
TopLevelDecl  = Declaration | FunctionDecl | MethodDecl .
</pre>

<p>
The <i>scope</i> of a declared identifier is the extent of source text in which
the identifier denotes the specified constant, type, variable, function, or package.
</p>

<p>
Go is lexically scoped using blocks:
</p>

<ol>
	<li>The scope of a predeclared identifier is the universe block.</li>

	<li>The scope of an identifier denoting a constant, type, variable,
	    or function declared at top level (outside any function) is the
	    package block.</li>

	<li>The scope of an imported package identifier is the file block
	    of the file containing the import declaration.</li>

	<li>The scope of an identifier denoting a function parameter or
	    result variable is the function body.</li>

	<li>The scope of a constant or variable identifier declared
	    inside a function begins at the end of the ConstSpec or VarSpec
	    and ends at the end of the innermost containing block.</li>

	<li>The scope of a type identifier declared inside a function
	    begins at the identifier in the TypeSpec
	    and ends at the end of the innermost containing block.</li>
</ol>

<p>
An identifier declared in a block may be redeclared in an inner block.
While the identifier of the inner declaration is in scope, it denotes
the entity declared by the inner declaration.
</p>

<p>
The <a href="#Package_clause">package clause</a> is not a declaration; the package name
does not appear in any scope. Its purpose is to identify the files belonging
to the same <a href="#Packages">package</a> and to specify the default package name for import
declarations.
</p>


<h3 id="Label_scopes">Label scopes</h3>

<p>
Labels are declared by <a href="#Labeled_statements">labeled statements</a> and are
used in the <code>break</code>, <code>continue</code>, and <code>goto</code>
statements (§<a href="#Break_statements">Break statements</a>, §<a href="#Continue_statements">Continue statements</a>, §<a href="#Goto_statements">Goto statements</a>).
In contrast to other identifiers, labels are not block scoped and do
not conflict with identifiers that are not labels. The scope of a label
is the body of the function in which it is declared and excludes
the body of any nested function.
</p>


<h3 id="Predeclared_identifiers">Predeclared identifiers</h3>

<p>
The following identifiers are implicitly declared in the universe block:
</p>
<pre class="grammar">
Basic types:
	bool byte float32 float64 int8 int16 int32 int64
	string uint8 uint16 uint32 uint64

Architecture-specific convenience types:
	float int uint uintptr

Constants:
	true false iota

Zero value:
	nil

Functions:
	cap close closed cmplx copy imag len make
	new panic print println real
</pre>


<h3 id="Exported_identifiers">Exported identifiers</h3>

<p>
An identifier may be <i>exported</i> to permit access to it from another package
using a <a href="#Qualified_identifiers">qualified identifier</a>. An identifier
is exported if both:
</p>
<ol>
	<li>the first character of the identifier's name is a Unicode upper case letter (Unicode class "Lu"); and</li>
	<li>the identifier is declared in the <a href="#Blocks">package block</a> or denotes a field or method of a type
	    declared in that block.</li>
</ol>
<p>
All other identifiers are not exported.
</p>


<h3 id="Blank_identifier">Blank identifier</h3>

<p>
The <i>blank identifier</i>, represented by the underscore character <code>_</code>, may be used in a declaration like
any other identifier but the declaration does not introduce a new binding.
</p>


<h3 id="Constant_declarations">Constant declarations</h3>

<p>
A constant declaration binds a list of identifiers (the names of
the constants) to the values of a list of <a href="#Constant_expressions">constant expressions</a>.
The number of identifiers must be equal
to the number of expressions, and the <i>n</i>th identifier on
the left is bound to the value of the <i>n</i>th expression on the
right.
</p>

<pre class="ebnf">
ConstDecl      = "const" ( ConstSpec | "(" { ConstSpec ";" } ")" ) .
ConstSpec      = IdentifierList [ [ Type ] "=" ExpressionList ] .

IdentifierList = identifier { "," identifier } .
ExpressionList = Expression { "," Expression } .
</pre>

<p>
If the type is present, all constants take the type specified, and
the expressions must be <a href="#Assignment_compatibility">assignment compatible</a> with that type.
If the type is omitted, the constants take the
individual types of the corresponding expressions.
If the expression values are untyped <a href="#Constants">constants</a>,
the declared constants remain untyped and the constant identifiers
denote the constant values. For instance, if the expression is a
floating-point literal, the constant identifier denotes a floating-point
constant, even if the literal's fractional part is zero.
</p>

<pre>
const Pi float64 = 3.14159265358979323846
const zero = 0.0             // untyped floating-point constant
const (
	size int64 = 1024
	eof = -1             // untyped integer constant
)
const a, b, c = 3, 4, "foo"  // a = 3, b = 4, c = "foo", untyped integer and string constants
const u, v float = 0, 3      // u = 0.0, v = 3.0
</pre>

<p>
Within a parenthesized <code>const</code> declaration list the
expression list may be omitted from any but the first declaration.
Such an empty list is equivalent to the textual substitution of the
first preceding non-empty expression list and its type if any.
Omitting the list of expressions is therefore equivalent to
repeating the previous list.  The number of identifiers must be equal
to the number of expressions in the previous list.
Together with the <a href="#Iota"><code>iota</code> constant generator</a>
this mechanism permits light-weight declaration of sequential values:
</p>

<pre>
const (
	Sunday = iota
	Monday
	Tuesday
	Wednesday
	Thursday
	Friday
	Partyday
	numberOfDays  // this constant is not exported
)
</pre>


<h3 id="Iota">Iota</h3>

<p>
Within a <a href="#Constant_declarations">constant declaration</a>, the predeclared identifier
<code>iota</code> represents successive untyped integer <a href="#Constants">
constants</a>. It is reset to 0 whenever the reserved word <code>const</code>
appears in the source and increments after each <a href="#ConstSpec">ConstSpec</a>.
It can be used to construct a set of related constants:
</p>

<pre>
const (  // iota is reset to 0
	c0 = iota  // c0 == 0
	c1 = iota  // c1 == 1
	c2 = iota  // c2 == 2
)

const (
	a = 1 &lt;&lt; iota  // a == 1 (iota has been reset)
	b = 1 &lt;&lt; iota  // b == 2
	c = 1 &lt;&lt; iota  // c == 4
)

const (
	u       = iota * 42  // u == 0     (untyped integer constant)
	v float = iota * 42  // v == 42.0  (float constant)
	w       = iota * 42  // w == 84    (untyped integer constant)
)

const x = iota  // x == 0 (iota has been reset)
const y = iota  // y == 0 (iota has been reset)
</pre>

<p>
Within an ExpressionList, the value of each <code>iota</code> is the same because
it is only incremented after each ConstSpec:
</p>

<pre>
const (
	bit0, mask0 = 1 &lt;&lt; iota, 1 &lt;&lt; iota - 1  // bit0 == 1, mask0 == 0
	bit1, mask1                             // bit1 == 2, mask1 == 1
	_, _                                    // skips iota == 2
	bit3, mask3                             // bit3 == 8, mask3 == 7
)
</pre>

<p>
This last example exploits the implicit repetition of the
last non-empty expression list.
</p>


<h3 id="Type_declarations">Type declarations</h3>

<p>
A type declaration binds an identifier, the <i>type name</i>, to a new type
that has the same definition (element, fields, channel direction, etc.) as
an existing type.  The new type is
<a href="#Properties_of_types_and_values">compatible</a> with, but
<a href="#Properties_of_types_and_values">different</a> from, the existing type.
</p>

<pre class="ebnf">
TypeDecl     = "type" ( TypeSpec | "(" { TypeSpec ";" } ")" ) .
TypeSpec     = identifier Type .
</pre>

<pre>
type IntArray [16]int

type (
	Point struct { x, y float }
	Polar Point
)

type TreeNode struct {
	left, right *TreeNode
	value *Comparable
}

type Cipher interface {
	BlockSize() int
	Encrypt(src, dst []byte)
	Decrypt(src, dst []byte)
}
</pre>

<p>
The declared type does not inherit any <a href="#Method_declarations">methods</a>
bound to the existing type, but the <a href="#Types">method set</a>
of an interface type or of elements of a composite type remains unchanged:
</p>

<pre>
// A Mutex is a data type with two methods Lock and Unlock.
type Mutex struct         { /* Mutex fields */ }
func (m *Mutex) Lock()    { /* Lock implementation */ }
func (m *Mutex) Unlock()  { /* Unlock implementation */ }

// NewMutex has the same composition as Mutex but its method set is empty.
type NewMutex Mutex

// The method set of *PrintableMutex contains the methods
// Lock and Unlock bound to its anonymous field Mutex.
type PrintableMutex struct {
	Mutex
}

// MyCipher is an interface type that has the same method set as Cipher.
type MyCipher Cipher
</pre>

<p>
A type declaration may be used to define a different boolean, numeric, or string
type and attach methods to it:
</p>

<pre>
type TimeZone int

const (
	EST TimeZone = -(5 + iota)
	CST
	MST
	PST
)

func (tz TimeZone) String() string {
	return fmt.Sprintf("GMT+%dh", tz)
}
</pre>


<h3 id="Variable_declarations">Variable declarations</h3>

<p>
A variable declaration creates a variable, binds an identifier to it and
gives it a type and optionally an initial value.
</p>
<pre class="ebnf">
VarDecl     = "var" ( VarSpec | "(" { VarSpec ";" } ")" ) .
VarSpec     = IdentifierList ( Type [ "=" ExpressionList ] | "=" ExpressionList ) .
</pre>

<pre>
var i int
var U, V, W float
var k = 0
var x, y float = -1, -2
var (
	i int
	u, v, s = 2.0, 3.0, "bar"
)
var re, im = complexSqrt(-1)
var _, found = entries[name]  // map lookup; only interested in "found"
</pre>

<p>
If a list of expressions is given, the variables are initialized
by assigning the expressions to the variables (§<a href="#Assignments">Assignments</a>)
in order; all expressions must be consumed and all variables initialized from them.
Otherwise, each variable is initialized to its <a href="#The_zero_value">zero value</a>.
</p>

<p>
If the type is present, each variable is given that type.
Otherwise, the types are deduced from the assignment
of the expression list.
</p>

<p>
If the type is absent and the corresponding expression evaluates to an
untyped <a href="#Constants">constant</a>, the type of the declared variable
is <code>bool</code>, <code>int</code>, <code>float</code>, or <code>string</code>
respectively, depending on whether the value is a boolean, integer,
floating-point, or string constant:
</p>

<pre>
var b = true    // t has type bool
var i = 0       // i has type int
var f = 3.0     // f has type float
var s = "OMDB"  // s has type string
</pre>

<h3 id="Short_variable_declarations">Short variable declarations</h3>

<p>
A <i>short variable declaration</i> uses the syntax:
</p>

<pre class="ebnf">
ShortVarDecl = IdentifierList ":=" ExpressionList .
</pre>

<p>
It is a shorthand for a regular variable declaration with
initializer expressions but no types:
</p>

<pre class="grammar">
"var" IdentifierList = ExpressionList .
</pre>

<pre>
i, j := 0, 10
f := func() int { return 7 }
ch := make(chan int)
r, w := os.Pipe(fd)  // os.Pipe() returns two values
_, y, _ := coord(p)  // coord() returns three values; only interested in y coordinate
</pre>

<p>
Unlike regular variable declarations, a short variable declaration may redeclare variables provided they
were originally declared in the same block with the same type, and at
least one of the non-<a href="#Blank_identifier">blank</a> variables is new.  As a consequence, redeclaration
can only appear in a multi-variable short declaration.
Redeclaration does not introduce a new
variable; it just assigns a new value to the original.
</p>

<pre>
field1, offset := nextField(str, 0)
field2, offset := nextField(str, offset)  // redeclares offset
</pre>

<p>
Short variable declarations may appear only inside functions.
In some contexts such as the initializers for <code>if</code>,
<code>for</code>, or <code>switch</code> statements,
they can be used to declare local temporary variables (§<a href="#Statements">Statements</a>).
</p>

<h3 id="Function_declarations">Function declarations</h3>

<p>
A function declaration binds an identifier to a function (§<a href="#Function_types">Function types</a>).
</p>

<pre class="ebnf">
FunctionDecl = "func" identifier Signature [ Body ] .
Body         = Block.
</pre>

<p>
A function declaration may omit the body. Such a declaration provides the
signature for a function implemented outside Go, such as an assembly routine.
</p>

<pre>
func min(x int, y int) int {
	if x &lt; y {
		return x
	}
	return y
}

func flushICache(begin, end uintptr)  // implemented externally
</pre>

<h3 id="Method_declarations">Method declarations</h3>

<p>
A method is a function with a <i>receiver</i>.
A method declaration binds an identifier to a method.
</p>
<pre class="ebnf">
MethodDecl   = "func" Receiver MethodName Signature [ Body ] .
Receiver     = "(" [ identifier ] [ "*" ] BaseTypeName ")" .
BaseTypeName = identifier .
</pre>

<p>
The receiver type must be of the form <code>T</code> or <code>*T</code> where
<code>T</code> is a type name. <code>T</code> is called the
<i>receiver base type</i> or just <i>base type</i>.
The base type must not be a pointer or interface type and must be
declared in the same package as the method.
The method is said to be <i>bound</i> to the base type
and is visible only within selectors for that type
(§<a href="#Type_declarations">Type declarations</a>, §<a href="#Selectors">Selectors</a>).
</p>

<p>
Given type <code>Point</code>, the declarations
</p>

<pre>
func (p *Point) Length() float {
	return Math.sqrt(p.x * p.x + p.y * p.y)
}

func (p *Point) Scale(factor float) {
	p.x = p.x * factor
	p.y = p.y * factor
}
</pre>

<p>
bind the methods <code>Length</code> and <code>Scale</code>,
with receiver type <code>*Point</code>,
to the base type <code>Point</code>.
</p>

<p>
If the receiver's value is not referenced inside the body of the method,
its identifier may be omitted in the declaration. The same applies in
general to parameters of functions and methods.
</p>

<p>
The type of a method is the type of a function with the receiver as first
argument.  For instance, the method <code>Scale</code> has type
</p>

<pre>
(p *Point, factor float)
</pre>

<p>
However, a function declared this way is not a method.
</p>


<h2 id="Expressions">Expressions</h2>

<p>
An expression specifies the computation of a value by applying
operators and functions to operands.
</p>

<h3 id="Operands">Operands</h3>

<p>
Operands denote the elementary values in an expression.
</p>

<pre class="ebnf">
Operand    = Literal | QualifiedIdent | MethodExpr | "(" Expression ")" .
Literal    = BasicLit | CompositeLit | FunctionLit .
BasicLit   = int_lit | float_lit | imaginary_lit | char_lit | string_lit .
</pre>


<h3 id="Qualified_identifiers">Qualified identifiers</h3>

<p>
A qualified identifier is a non-<a href="#Blank_identifier">blank</a> identifier qualified by a package name prefix.
</p>

<pre class="ebnf">
QualifiedIdent = [ PackageName "." ] identifier .
</pre>

<p>
A qualified identifier accesses an identifier in a separate package.
The identifier must be <a href="#Exported_identifiers">exported</a> by that
package, which means that it must begin with a Unicode upper case letter.
</p>

<pre>
math.Sin
</pre>

<!---
<p>
<span class="alert">TODO: Unify this section with Selectors - it's the same syntax.</span>
</p>
--->

<h3 id="Composite_literals">Composite literals</h3>

<p>
Composite literals construct values for structs, arrays, slices, and maps
and create a new value each time they are evaluated.
They consist of the type of the value
followed by a brace-bound list of composite elements. An element may be
a single expression or a key-value pair.
</p>

<pre class="ebnf">
CompositeLit  = LiteralType "{" [ ElementList [ "," ] ] "}" .
LiteralType   = StructType | ArrayType | "[" "..." "]" ElementType |
                SliceType | MapType | TypeName | "(" LiteralType ")" .
ElementList   = Element { "," Element } .
Element       = [ Key ":" ] Value .
Key           = FieldName | ElementIndex .
FieldName     = identifier .
ElementIndex  = Expression .
Value         = Expression .
</pre>

<p>
The LiteralType must be a struct, array, slice, or map type
(the grammar enforces this constraint except when the type is given
as a TypeName).
The types of the expressions must be <a href="#Assignment_compatibility">assignment compatible</a> with
the respective field, element, and key types of the LiteralType;
there is no additional conversion.
The key is interpreted as a field name for struct literals,
an index expression for array and slice literals, and a key for map literals.
For map literals, all elements must have a key. It is an error
to specify multiple elements with the same field name or
constant key value.
</p>

<p>
For struct literals the following rules apply:
</p>
<ul>
	<li>A key must be a field name declared in the LiteralType.
	</li>
	<li>A literal that does not contain any keys must
	    list an element for each struct field in the
	    order in which the fields are declared.
	</li>
	<li>If any element has a key, every element must have a key.
	</li>
	<li>A literal that contains keys does not need to
	    have an element for each struct field. Omitted fields
	    get the zero value for that field.
	</li>
	<li>A literal may omit the element list; such a literal evaluates
		to the zero value for its type.
	</li>
	<li>It is an error to specify an element for a non-exported
	    field of a struct belonging to a different package.
	</li>
</ul>

<p>
Given the declarations
</p>
<pre>
type Point struct { x, y, z float }
type Line struct { p, q Point }
</pre>

<p>
one may write
</p>

<pre>
origin := Point{}                            // zero value for Point
line := Line{origin, Point{y: -4, z: 12.3}}  // zero value for line.q.x
</pre>

<p>
For array and slice literals the following rules apply:
</p>
<ul>
	<li>Each element has an associated integer index marking
	    its position in the array.
	</li>
	<li>An element with a key uses the key as its index; the
	    key must be a constant integer expression.
	</li>
	<li>An element without a key uses the previous element's index plus one.
	    If the first element has no key, its index is zero.
	</li>
</ul>

<p>
Taking the address of a composite literal (§<a href="#Address_operators">Address operators</a>)
generates a unique pointer to an instance of the literal's value.
</p>
<pre>
var pointer *Point = &amp;Point{y: 1000}
</pre>

<p>
The length of an array literal is the length specified in the LiteralType.
If fewer elements than the length are provided in the literal, the missing
elements are set to the zero value for the array element type.
It is an error to provide elements with index values outside the index range
of the array. The notation <code>...</code> specifies an array length equal
to the maximum element index plus one.
</p>

<pre>
buffer := [10]string{}               // len(buffer) == 10
intSet := [6]int{1, 2, 3, 5}         // len(intSet) == 6
days := [...]string{"Sat", "Sun"}    // len(days) == 2
</pre>

<p>
A slice literal describes the entire underlying array literal.
Thus, the length and capacity of a slice literal are the maximum
element index plus one. A slice literal has the form
</p>

<pre>
[]T{x1, x2, ... xn}
</pre>

<p>
and is a shortcut for a slice operation applied to an array literal:
</p>

<pre>
[n]T{x1, x2, ... xn}[0 : n]
</pre>

<p>
A parsing ambiguity arises when a composite literal using the
TypeName form of the LiteralType appears in the condition of an
"if", "for", or "switch" statement, because the braces surrounding
the expressions in the literal are confused with those introducing
a block of statements. To resolve the ambiguity in this rare case,
the composite literal must appear within
parentheses.
</p>

<pre>
if x == (T{a,b,c}[i]) { ... }
if (x == T{a,b,c}[i]) { ... }
</pre>

<p>
Examples of valid array, slice, and map literals:
</p>

<pre>
// list of prime numbers
primes := []int{2, 3, 5, 7, 9, 11, 13, 17, 19, 991}

// vowels[ch] is true if ch is a vowel
vowels := [128]bool{'a': true, 'e': true, 'i': true, 'o': true, 'u': true, 'y': true}

// the array [10]float{-1, 0, 0, 0, -0.1, -0.1, 0, 0, 0, -1}
filter := [10]float{-1, 4: -0.1, -0.1, 9: -1}

// frequencies in Hz for equal-tempered scale (A4 = 440Hz)
noteFrequency := map[string]float{
	"C0": 16.35, "D0": 18.35, "E0": 20.60, "F0": 21.83,
	"G0": 24.50, "A0": 27.50, "B0": 30.87,
}
</pre>


<h3 id="Function_literals">Function literals</h3>

<p>
A function literal represents an anonymous function.
It consists of a specification of the function type and a function body.
</p>

<pre class="ebnf">
FunctionLit = FunctionType Body .
</pre>

<pre>
func(a, b int, z float) bool { return a*b &lt; int(z) }
</pre>

<p>
A function literal can be assigned to a variable or invoked directly.
</p>

<pre>
f := func(x, y int) int { return x + y }
func(ch chan int) { ch &lt;- ACK } (reply_chan)
</pre>

<p>
Function literals are <i>closures</i>: they may refer to variables
defined in a surrounding function. Those variables are then shared between
the surrounding function and the function literal, and they survive as long
as they are accessible.
</p>


<h3 id="Primary_expressions">Primary expressions</h3>

<p>
Primary expressions are the operands for unary and binary expressions.
</p>

<pre class="ebnf">
PrimaryExpr =
	Operand |
	Conversion |
	BuiltinCall |
	PrimaryExpr Selector |
	PrimaryExpr Index |
	PrimaryExpr Slice |
	PrimaryExpr TypeAssertion |
	PrimaryExpr Call .

Selector       = "." identifier .
Index          = "[" Expression "]" .
Slice          = "[" Expression ":" [ Expression ] "]" .
TypeAssertion  = "." "(" Type ")" .
Call           = "(" [ ExpressionList [ "," ] ] ")" .
</pre>


<pre>
x
2
(s + ".txt")
f(3.1415, true)
Point{1, 2}
m["foo"]
s[i : j + 1]
obj.color
Math.sin
f.p[i].x()
</pre>


<h3 id="Selectors">Selectors</h3>

<p>
A primary expression of the form
</p>

<pre>
x.f
</pre>

<p>
denotes the field or method <code>f</code> of the value denoted by <code>x</code>
(or of <code>*x</code> if
<code>x</code> is of pointer type). The identifier <code>f</code>
is called the (field or method)
<i>selector</i>; it must not be the <a href="#Blank_identifier">blank identifier</a>.
The type of the expression is the type of <code>f</code>.
</p>
<p>
A selector <code>f</code> may denote a field or method <code>f</code> of
a type <code>T</code>, or it may refer
to a field or method <code>f</code> of a nested anonymous field of
<code>T</code>.
The number of anonymous fields traversed
to reach <code>f</code> is called its <i>depth</i> in <code>T</code>.
The depth of a field or method <code>f</code>
declared in <code>T</code> is zero.
The depth of a field or method <code>f</code> declared in
an anonymous field <code>A</code> in <code>T</code> is the
depth of <code>f</code> in <code>A</code> plus one.
</p>
<p>
The following rules apply to selectors:
</p>
<ol>
<li>
For a value <code>x</code> of type <code>T</code> or <code>*T</code>
where <code>T</code> is not an interface type,
<code>x.f</code> denotes the field or method at the shallowest depth
in <code>T</code> where there
is such an <code>f</code>.
If there is not exactly one <code>f</code> with shallowest depth, the selector
expression is illegal.
</li>
<li>
For a variable <code>x</code> of type <code>I</code> or <code>*I</code>
where <code>I</code> is an interface type,
<code>x.f</code> denotes the actual method with name <code>f</code> of the value assigned
to <code>x</code> if there is such a method.
If no value or <code>nil</code> was assigned to <code>x</code>, <code>x.f</code> is illegal.
</li>
<li>
In all other cases, <code>x.f</code> is illegal.
</li>
</ol>
<p>
Selectors automatically dereference pointers.
If <code>x</code> is of pointer type, <code>x.y</code>
is shorthand for <code>(*x).y</code>; if <code>y</code>
is also of pointer type, <code>x.y.z</code> is shorthand
for <code>(*(*x).y).z</code>, and so on.
If <code>*x</code> is of pointer type, dereferencing
must be explicit;
only one level of automatic dereferencing is provided.
For an <code>x</code> of type <code>T</code> containing an
anonymous field declared as <code>*A</code>,
<code>x.f</code> is a shortcut for <code>(*x.A).f</code>.
</p>
<p>
For example, given the declarations:
</p>

<pre>
type T0 struct {
	x int
}

func (recv *T0) M0()

type T1 struct {
	y int
}

func (recv T1) M1()

type T2 struct {
	z int
	T1
	*T0
}

func (recv *T2) M2()

var p *T2  // with p != nil and p.T1 != nil
</pre>

<p>
one may write:
</p>

<pre>
p.z         // (*p).z
p.y         // ((*p).T1).y
p.x         // (*(*p).T0).x

p.M2        // (*p).M2
p.M1        // ((*p).T1).M1
p.M0        // ((*p).T0).M0
</pre>


<!---
<span class="alert">
TODO: Specify what happens to receivers.
</span>
--->


<h3 id="Indexes">Indexes</h3>

<p>
A primary expression of the form
</p>

<pre>
a[x]
</pre>

<p>
denotes the element of the array, slice, string or map <code>a</code> indexed by <code>x</code>.
The value <code>x</code> is called the
<i>index</i> or <i>map key</i>, respectively. The following
rules apply:
</p>

<p>
For <code>a</code> of type <code>A</code> or <code>*A</code>
where <code>A</code> is an <a href="#Array_types">array type</a>,
or for <code>a</code> of type <code>S</code> where <code>S</code> is a <a href="#Slice_types">slice type</a>:
</p>
<ul>
	<li><code>x</code> must be an integer value and <code>0 &lt;= x &lt; len(a)</code></li>
	<li><code>a[x]</code> is the array element at index <code>x</code> and the type of
	  <code>a[x]</code> is the element type of <code>A</code></li>
	<li>if the index <code>x</code> is out of range,
	a <a href="#Run_time_panics">run-time panic</a> occurs</li>
</ul>

<p>
For <code>a</code> of type <code>T</code>
where <code>T</code> is a <a href="#String_types">string type</a>:
</p>
<ul>
	<li><code>x</code> must be an integer value and <code>0 &lt;= x &lt; len(a)</code></li>
	<li><code>a[x]</code> is the byte at index <code>x</code> and the type of
	  <code>a[x]</code> is <code>byte</code></li>
	<li><code>a[x]</code> may not be assigned to</li>
	<li>if the index <code>x</code> is out of range,
	a <a href="#Run_time_panics">run-time panic</a> occurs</li>
</ul>

<p>
For <code>a</code> of type <code>M</code>
where <code>M</code> is a <a href="#Map_types">map type</a>:
</p>
<ul>
	<li><code>x</code>'s type must be
	<a href="#Assignment_compatibility">assignment compatible</a>
	with the key type of <code>M</code></li>
	<li>if the map contains an entry with key <code>x</code>,
	  <code>a[x]</code> is the map value with key <code>x</code>
	  and the type of <code>a[x]</code> is the value type of <code>M</code></li>
	<li>if the map does not contain such an entry,
	  <code>a[x]</code> is the <a href="#The_zero_value">zero value</a>
	  for the value type of <code>M</code></li>
</ul>

<p>
Otherwise <code>a[x]</code> is illegal.
</p>

<p>
An index expression on a map <code>a</code> of type <code>map[K]V</code>
may be used in an assignment or initialization of the special form
</p>

<pre>
v, ok = a[x]
v, ok := a[x]
var v, ok = a[x]
</pre>

<p>
where the result of the index expression is a pair of values with types
<code>(V, bool)</code>. In this form, the value of <code>ok</code> is
<code>true</code> if the key <code>x</code> is present in the map, and
<code>false</code> otherwise. The value of <code>v</code> is the value
<code>a[x]</code> as in the single-result form.
</p>

<p>
Similarly, if an assignment to a map has the special form
</p>

<pre>
a[x] = v, ok
</pre>

<p>
and boolean <code>ok</code> has the value <code>false</code>,
the entry for key <code>x</code> is deleted from the map; if
<code>ok</code> is <code>true</code>, the construct acts like
a regular assignment to an element of the map.
</p>


<h3 id="Slices">Slices</h3>

<p>
For a string, array, or slice <code>a</code>, the primary expression
</p>

<pre>
a[lo : hi]
</pre>

<p>
constructs a substring or slice. The index expressions <code>lo</code> and
<code>hi</code> select which elements appear in the result. The result has
indexes starting at 0 and length equal to
<code>hi</code>&nbsp;-&nbsp;<code>lo</code>.
After slicing the array <code>a</code>
</p>

<pre>
a := [5]int{1, 2, 3, 4, 5}
s := a[1:4]
</pre>

<p>
the slice <code>s</code> has type <code>[]int</code>, length 3, capacity 4, and elements
</p>

<pre>
s[0] == 2
s[1] == 3
s[2] == 4
</pre>

<p>
For convenience, the <code>hi</code> expression may be omitted; the notation
<code>a[lo :]</code> is shorthand for <code>a[lo : len(a)]</code>.
For arrays or strings, the indexes
<code>lo</code> and <code>hi</code> must satisfy
0 &lt;= <code>lo</code> &lt;= <code>hi</code> &lt;= length;
for slices, the upper bound is the capacity rather than the length.
</p>

<p>
If the sliced operand is a string or slice, the result of the slice operation
is a string or slice of the same type.
If the sliced operand is an array, the result of the slice operation is a slice
with the same element type as the array.
</p>


<h3 id="Type_assertions">Type assertions</h3>

<p>
For an expression <code>x</code> of <a href="#Interface_types">interface type</a>
and a type <code>T</code>, the primary expression
</p>

<pre>
x.(T)
</pre>

<p>
asserts that <code>x</code> is not <code>nil</code>
and that the value stored in <code>x</code> is of type <code>T</code>.
The notation <code>x.(T)</code> is called a <i>type assertion</i>.
</p>
<p>
More precisely, if <code>T</code> is not an interface type, <code>x.(T)</code> asserts
that the dynamic type of <code>x</code> is identical to the type <code>T</code>
(§<a href="#Type_identity">Type identity and compatibility</a>).
If <code>T</code> is an interface type, <code>x.(T)</code> asserts that the dynamic type
of <code>x</code> implements the interface <code>T</code> (§<a href="#Interface_types">Interface types</a>).
</p>
<p>
If the type assertion holds, the value of the expression is the value
stored in <code>x</code> and its type is <code>T</code>. If the type assertion is false,
a <a href="#Run_time_panics">run-time panic</a> occurs.
In other words, even though the dynamic type of <code>x</code>
is known only at run-time, the type of <code>x.(T)</code> is
known to be <code>T</code> in a correct program.
</p>
<p>
If a type assertion is used in an assignment or initialization of the form
</p>

<pre>
v, ok = x.(T)
v, ok := x.(T)
var v, ok = x.(T)
</pre>

<p>
the result of the assertion is a pair of values with types <code>(T, bool)</code>.
If the assertion holds, the expression returns the pair <code>(x.(T), true)</code>;
otherwise, the expression returns <code>(Z, false)</code> where <code>Z</code>
is the <a href="#The_zero_value">zero value</a> for type <code>T</code>.
No run-time panic occurs in this case.
The type assertion in this construct thus acts like a function call
returning a value and a boolean indicating success.  (§<a href="#Assignments">Assignments</a>)
</p>


<h3 id="Calls">Calls</h3>

<p>
Given an expression <code>f</code> of function type
<code>F</code>,
</p>

<pre>
f(a1, a2, ... an)
</pre>

<p>
calls <code>f</code> with arguments <code>a1, a2, ... an</code>.
Except for one special case, arguments must be single-valued expressions
<a href="#Assignment_compatibility">assignment compatible</a> with the parameter types of
<code>F</code> and are evaluated before the function is called.
The type of the expression is the result type
of <code>F</code>.
A method invocation is similar but the method itself
is specified as a selector upon a value of the receiver type for
the method.
</p>

<pre>
math.Atan2(x, y)    // function call
var pt *Point
pt.Scale(3.5)  // method call with receiver pt
</pre>

<p>
As a special case, if the return parameters of a function or method
<code>g</code> are equal in number and individually assignment
compatible with the parameters of another function or method
<code>f</code>, then the call <code>f(g(<i>parameters_of_g</i>))</code>
will invoke <code>f</code> after binding the return values of
<code>g</code> to the parameters of <code>f</code> in order.  The call
of <code>f</code> must contain no parameters other than the call of <code>g</code>.
If <code>f</code> has a final <code>...</code> parameter, it is
assigned the return values of <code>g</code> that remain after
assignment of regular parameters.
</p>

<pre>
func Split(s string, pos int) (string, string) {
	return s[0:pos], s[pos:]
}

func Join(s, t string) string {
	return s + t
}

if Join(Split(value, len(value)/2)) != value {
	log.Crash("test fails")
}
</pre>

<p>
A method call <code>x.m()</code> is valid if the method set of
(the type of) <code>x</code> contains <code>m</code> and the
argument list can be assigned to the parameter list of <code>m</code>.
If <code>x</code> is <a href="#Address_operators">addressable</a> and <code>&amp;x</code>'s method
set contains <code>m</code>, <code>x.m()</code> is shorthand
for <code>(&amp;x).m()</code>:
</p>

<pre>
var p Point
p.Scale(3.5)
</pre>

<p>
There is no distinct method type and there are no method literals.
</p>

<h3 id="Passing_arguments_to_..._parameters">Passing arguments to <code>...</code> parameters</h3>

<p>
When a function <code>f</code> has a <code>...</code> parameter,
it is always the last formal parameter. Within calls to <code>f</code>,
the arguments before the <code>...</code> are treated normally.
After those, an arbitrary number (including zero) of trailing
arguments may appear in the call and are bound to the <code>...</code>
parameter.
</p>

<p>
Within <code>f</code>, a <code>...</code> parameter with no
specified type has static type <code>interface{}</code> (the empty
interface). For each call, its dynamic type is a structure whose
sequential fields are the trailing arguments of the call.  That is,
the actual arguments provided for a <code>...</code> parameter are
wrapped into a struct that is passed to the function instead of the
actual arguments.  Using the <a href="#Package_unsafe">reflection</a>
interface, <code>f</code> may unpack the elements of the dynamic
type to recover the actual arguments.
</p>

<p>
Given the function and call
</p>
<pre>
func Fprintf(f io.Writer, format string, args ...)
Fprintf(os.Stdout, "%s %d", "hello", 23)
</pre>

<p>
Within <code>Fprintf</code>, the dynamic type of <code>args</code> for this
call will be, schematically,
<code> struct { string; int }</code>.
</p>

<p>
If the final parameter of <code>f</code> has type <code>... T</code>,
within the function it is equivalent to a parameter of type
<code>[]T</code>.  At each call of <code>f</code>, the actual
arguments provided for the <code>... T</code> parameter are placed
into a new slice of type <code>[]T</code> whose successive elements are
the actual arguments.  The length of the slice is therefore the
number of arguments bound to the <code>... T</code> parameter and
may differ for each call site.
</p>

<p>
Given the function and call
</p>
<pre>
func Greeting(prefix string, who ... string)
Greeting("hello:", "Joe", "Anna", "Eileen")
</pre>

<p>
Within <code>Greeting</code>, <code>who</code> will have value
<code>[]string{"Joe", "Anna", "Eileen")</code>
</p>


<p>
As a special case, if a function passes its own <code>...</code> parameter,
with or without specified type, as the argument
for a <code>...</code> in a call to another function with a <code>...</code> parameter
of identical type,
the parameter is not wrapped again but passed directly. In short, a formal <code>...</code>
parameter is passed unchanged as an actual <code>...</code> parameter provided the
types match.
</p>

<h3 id="Operators">Operators</h3>

<p>
Operators combine operands into expressions.
</p>

<pre class="ebnf">
Expression = UnaryExpr | Expression binary_op UnaryExpr .
UnaryExpr  = PrimaryExpr | unary_op UnaryExpr .

binary_op  = log_op | com_op | rel_op | add_op | mul_op .
log_op     = "||" | "&amp;&amp;" .
com_op     = "&lt;-" .
rel_op     = "==" | "!=" | "&lt;" | "&lt;=" | ">" | ">=" .
add_op     = "+" | "-" | "|" | "^" .
mul_op     = "*" | "/" | "%" | "&lt;&lt;" | "&gt;&gt;" | "&amp;" | "&amp;^" .

unary_op   = "+" | "-" | "!" | "^" | "*" | "&amp;" | "&lt;-" .
</pre>

<p>
Comparisons are discussed <a href="#Comparison_operators">elsewhere</a>.
For other binary operators, the operand types must be identical
(§<a href="#Properties_of_types_and_values">Properties of types and values</a>)
unless the operation involves channels, shifts, or untyped <a href="#Constants">constants</a>.
For operations involving constants only, see the section on
<a href="#Constant_expressions">constant expressions</a>.
</p>

<p>
In a channel send, the first operand is always a channel and the second
must be a value <a href="#Assignment_compatibility">assignment compatible</a>
with the channel's element type.
</p>

<p>
Except for shift operations,
if one operand is an untyped <a href="#Constants">constant</a>
and the other operand is not, the constant is <a href="#Conversions">converted</a>
to the type of the other operand.
</p>

<p>
The right operand in a shift operation must have unsigned integer type
or be an untyped constant that can be converted to unsigned integer type.
</p>

<p>
If the left operand of a non-constant shift operation is an untyped constant,
the type of constant is what it would be if the shift operation were replaced by
the left operand alone.
</p>

<pre>
var s uint = 33
var i = 1&lt;&lt;s          // 1 has type int
var j = int32(1&lt;&lt;s)   // 1 has type int32; j == 0
var u = uint64(1&lt;&lt;s)  // 1 has type uint64; u == 1&lt;&lt;33
var f = float(1&lt;&lt;s)   // illegal: 1 has type float, cannot shift
var g = float(1&lt;&lt;33)  // legal; 1&lt;&lt;33 is a constant shift operation; g == 1&lt;&lt;33
</pre>

<h3 id="Operator_precedence">Operator precedence</h3>
<p>
Unary operators have the highest precedence.
As the  <code>++</code> and <code>--</code> operators form
statements, not expressions, they fall
outside the operator hierarchy.
As a consequence, statement <code>*p++</code> is the same as <code>(*p)++</code>.
<p>
There are six precedence levels for binary operators.
Multiplication operators bind strongest, followed by addition
operators, comparison operators, <code>&lt;-</code> (channel send),
<code>&amp;&amp;</code> (logical and), and finally <code>||</code> (logical or):
</p>

<pre class="grammar">
Precedence    Operator
    6             *  /  %  &lt;&lt;  &gt;&gt;  &amp;  &amp;^
    5             +  -  |  ^
    4             ==  !=  &lt;  &lt;=  >  >=
    3             &lt;-
    2             &amp;&amp;
    1             ||
</pre>

<p>
Binary operators of the same precedence associate from left to right.
For instance, <code>x / y * z</code> is the same as <code>(x / y) * z</code>.
</p>

<pre>
+x
23 + 3*x[i]
x &lt;= f()
^a &gt;&gt; b
f() || g()
x == y+1 &amp;&amp; &lt;-chan_ptr > 0
</pre>


<h3 id="Arithmetic_operators">Arithmetic operators</h3>
<p>
Arithmetic operators apply to numeric values and yield a result of the same
type as the first operand. The four standard arithmetic operators (<code>+</code>,
<code>-</code>,  <code>*</code>, <code>/</code>) apply to integer,
floating-point, and complex types; <code>+</code> also applies
to strings. All other arithmetic operators apply to integers only.
</p>

<pre class="grammar">
+    sum                    integers, floats, complex values, strings
-    difference             integers, floats, complex values
*    product                integers, floats, complex values
/    quotient               integers, floats, complex values
%    remainder              integers

&amp;    bitwise and            integers
|    bitwise or             integers
^    bitwise xor            integers
&amp;^   bit clear (and not)    integers

&lt;&lt;   left shift             integer &lt;&lt; unsigned integer
&gt;&gt;   right shift            integer &gt;&gt; unsigned integer
</pre>

<p>
Strings can be concatenated using the <code>+</code> operator
or the <code>+=</code> assignment operator:
</p>

<pre>
s := "hi" + string(c)
s += " and good bye"
</pre>

<p>
String addition creates a new string by concatenating the operands.
</p>
<p>
For integer values, <code>/</code> and <code>%</code> satisfy the following relationship:
</p>

<pre>
(a / b) * b + a % b == a
</pre>

<p>
with <code>(a / b)</code> truncated towards zero.
</p>

<pre>
 x     y     x / y     x % y
 5     3       1         2
-5     3      -1        -2
 5    -3      -1         2
-5    -3       1        -2
</pre>

<p>
If the divisor is zero, a <a href="#Run_time_panics">run-time panic</a> occurs.
If the dividend is positive and the divisor is a constant power of 2,
the division may be replaced by a right shift, and computing the remainder may
be replaced by a bitwise "and" operation:
</p>

<pre>
 x     x / 4     x % 4     x &gt;&gt; 2     x &amp; 3
 11      2         3         2          3
-11     -2        -3        -3          1
</pre>

<p>
The shift operators shift the left operand by the shift count specified by the
right operand. They implement arithmetic shifts if the left operand is a signed
integer and logical shifts if it is an unsigned integer. The shift count must
be an unsigned integer. There is no upper limit on the shift count. Shifts behave
as if the left operand is shifted <code>n</code> times by 1 for a shift
count of <code>n</code>.
As a result, <code>x &lt;&lt; 1</code> is the same as <code>x*2</code>
and <code>x &gt;&gt; 1</code> is the same as
<code>x/2</code> but truncated towards negative infinity.
</p>

<p>
For integer operands, the unary operators
<code>+</code>, <code>-</code>, and <code>^</code> are defined as
follows:
</p>

<pre class="grammar">
+x                          is 0 + x
-x    negation              is 0 - x
^x    bitwise complement    is m ^ x  with m = "all bits set to 1" for unsigned x
                                      and  m = -1 for signed x
</pre>

<p>
For floating-point numbers,
<code>+x</code> is the same as <code>x</code>,
while <code>-x</code> is the negation of <code>x</code>.
The result of a floating-point division by zero is not specified beyond the
IEEE-754 standard; whether a <a href="#Run_time_panics">run-time panic</a>
occurs is implementation-specific.
</p>

<h3 id="Integer_overflow">Integer overflow</h3>

<p>
For unsigned integer values, the operations <code>+</code>,
<code>-</code>, <code>*</code>, and <code>&lt;&lt;</code> are
computed modulo 2<sup><i>n</i></sup>, where <i>n</i> is the bit width of
the unsigned integer's type
(§<a href="#Numeric_types">Numeric types</a>). Loosely speaking, these unsigned integer operations
discard high bits upon overflow, and programs may rely on ``wrap around''.
</p>
<p>
For signed integers, the operations <code>+</code>,
<code>-</code>, <code>*</code>, and <code>&lt;&lt;</code> may legally
overflow and the resulting value exists and is deterministically defined
by the signed integer representation, the operation, and its operands.
No exception is raised as a result of overflow. A
compiler may not optimize code under the assumption that overflow does
not occur. For instance, it may not assume that <code>x &lt; x + 1</code> is always true.
</p>


<h3 id="Comparison_operators">Comparison operators</h3>

<p>
Comparison operators compare two operands and yield a value of type <code>bool</code>.
</p>

<pre class="grammar">
==    equal
!=    not equal
<     less
<=    less or equal
>     greater
>=    greater or equal
</pre>

<p>
The operands must be <i>comparable</i>; that is, the first operand
must be <a href="#Assignment_compatibility">assignment compatible</a>
with the type of the second operand, or vice versa.
<p>
</p>
The operators <code>==</code> and <code>!=</code> apply
to operands of all types except arrays and structs.
All other comparison operators apply only to integer, floating-point
and string values. The result of a comparison is defined as follows:
</p>

<ul>
	<li>
	Integer values are compared in the usual way.
	</li>
	<li>
	Floating point values are compared as defined by the IEEE-754
	standard.
	</li>
	<li>
	Two complex values <code>u</code>, <code>v</code> are
	equal if both <code>real(u) == real(v)</code> and
	<code>imag(u) == imag(v)</code>.
	</li>
	<li>
	String values are compared byte-wise (lexically).
	</li>
	<li>
	Boolean values are are equal if they are either both
	<code>true</code> or both <code>false</code>.
	</li>
	<li>
	Pointer values are equal if they point to the same location
	or if both are <code>nil</code>.
	</li>
	<li>
	Function values are equal if they refer to the same function
	or if both are <code>nil</code>.
	</li>
	<li>
	A slice value may only be compared to <code>nil</code>.
	</li>
	<li>
	Channel and map values are equal if they were created by the same call to <code>make</code>
	(§<a href="#Making_slices_maps_and_channels">Making slices, maps, and channels</a>)
	or if both are <code>nil</code>.
	</li>
	<li>
	Interface values are equal if they have identical dynamic types and
	equal dynamic values or if both are <code>nil</code>.
	</li>
	<li>
	An interface value <code>x</code> is equal to a non-interface value
	<code>y</code> if the dynamic type of <code>x</code> is identical to
	the static type of <code>y</code> and the dynamic value of <code>x</code>
	is equal to <code>y</code>.
	</li>
	<li>
	A pointer, function, slice, channel, map, or interface value is equal
	to <code>nil</code> if it has been assigned the explicit value
	<code>nil</code>, if it is uninitialized, or if it has been assigned
	another value equal to <code>nil</code>.
	</li>
</ul>


<h3 id="Logical_operators">Logical operators</h3>

<p>
Logical operators apply to <a href="#Boolean_types">boolean</a> values
and yield a result of the same type as the operands.
The right operand is evaluated conditionally.
</p>

<pre class="grammar">
&amp;&amp;    conditional and    p &amp;&amp; q  is  "if p then q else false"
||    conditional or     p || q  is  "if p then true else q"
!     not                !p      is  "not p"
</pre>


<h3 id="Address_operators">Address operators</h3>

<p>
The address-of operator <code>&amp;</code> generates the address of its operand,
which must be <i>addressable</i>,
that is, either a variable, pointer indirection, or slice indexing
operation;
or a field selector of an addressable struct operand;
or an array indexing operation of an addressable array.
Given an operand of pointer type, the pointer indirection
operator <code>*</code> retrieves the value pointed
to by the operand.
</p>

<pre>
&amp;x
&amp;a[f(2)]
*p
*pf(x)
</pre>

<h3 id="Communication_operators">Communication operators</h3>

<p>
The term <i>channel</i> means "value of <a href="#Channel_types">channel type</a>".
</p>
<p>
The send operation uses the binary operator "&lt;-", which operates on
a channel and a value (expression):
</p>

<pre>
ch &lt;- 3
</pre>

<p>
The send operation sends the value on the channel.  Both the channel
and the expression are evaluated before communication begins.
Communication blocks until the send can proceed, at which point the
value is transmitted on the channel.
A send on an unbuffered channel can proceed if a receiver is ready.
A send on a buffered channel can proceed if there is room in the buffer.
</p>
<p>
If the send operation appears in an expression context, the value
of the expression is a boolean and the operation is non-blocking.
The value of the boolean reports true if the communication succeeded,
false if it did not. (The channel and
the expression to be sent are evaluated regardless.)
These two examples are equivalent:
</p>

<pre>
ok := ch &lt;- 3
if ok { print("sent") } else { print("not sent") }

if ch &lt;- 3 { print("sent") } else { print("not sent") }
</pre>

<p>
In other words, if the program tests the value of a send operation,
the send is non-blocking and the value of the expression is the
success of the operation.  If the program does not test the value,
the operation blocks until it succeeds.
</p>
<p>
The receive operation uses the prefix unary operator "&lt;-".
The value of the expression is the value received, whose type
is the element type of the channel.
</p>

<pre>
&lt;-ch
</pre>

<p>
The expression blocks until a value is available, which then can
be assigned to a variable or used like any other expression.
If the receive expression does not save the value, the value is
discarded.
</p>

<pre>
v1 := &lt;-ch
v2 = &lt;-ch
f(&lt;-ch)
&lt;-strobe  // wait until clock pulse
</pre>

<p>
If a receive expression is used in an assignment or initialization of the form
</p>

<pre>
x, ok = &lt;-ch
x, ok := &lt;-ch
var x, ok = &lt;-ch
</pre>

<p>
the receive operation becomes non-blocking.
If the operation can proceed, the boolean variable
<code>ok</code> will be set to <code>true</code>
and the value stored in <code>x</code>; otherwise
<code>ok</code> is set
to <code>false</code> and <code>x</code> is set to the
zero value for its type (§<a href="#The_zero_value">The zero value</a>).
</p>

<!---
<p>
<span class="alert">TODO: Probably in a separate section, communication semantics
need to be presented regarding send, receive, select, and goroutines.</span>
</p>
--->

<h3 id="Method_expressions">Method expressions</h3>

<p>
If <code>M</code> is in the method set of type <code>T</code>,
<code>T.M</code> is a function that is callable as a regular function
with the same arguments as <code>M</code> prefixed by an additional
argument that is the receiver of the method.
</p>

<pre class="ebnf">
MethodExpr    = ReceiverType "." MethodName .
ReceiverType  = TypeName | "(" "*" TypeName ")" .
</pre>

<p>
Consider a struct type <code>T</code> with two methods,
<code>Mv</code>, whose receiver is of type <code>T</code>, and
<code>Mp</code>, whose receiver is of type <code>*T</code>.
</p>

<pre>
type T struct {
	a int
}
func (tv  T) Mv(a int)   int   { return 0 }  // value receiver
func (tp *T) Mp(f float) float { return 1 }  // pointer receiver
var t T
</pre>

<p>
The expression
</p>

<pre>
T.Mv
</pre>

<p>
yields a function equivalent to <code>Mv</code> but
with an explicit receiver as its first argument; it has signature
</p>

<pre>
func(tv T, a int) int
</pre>

<p>
That function may be called normally with an explicit receiver, so
these three invocations are equivalent:
</p>

<pre>
t.Mv(7)
T.Mv(t, 7)
f := T.Mv; f(t, 7)
</pre>

<p>
Similarly, the expression
</p>

<pre>
(*T).Mp
</pre>

<p>
yields a function value representing <code>Mp</code> with signature
</p>

<pre>
func(tp *T, f float) float
</pre>

<p>
For a method with a value receiver, one can derive a function
with an explicit pointer receiver, so
</p>

<pre>
(*T).Mv
</pre>

<p>
yields a function value representing <code>Mv</code> with signature
</p>

<pre>
func(tv *T, a int) int
</pre>

<p>
Such a function indirects through the receiver to create a value
to pass as the receiver to the underlying method;
the method does not overwrite the value whose address is passed in
the function call.
</p>

<p>
The final case, a value-receiver function for a pointer-receiver method,
is illegal because pointer-receiver methods are not in the method set
of the value type.
</p>

<p>
Function values derived from methods are called with function call syntax;
the receiver is provided as the first argument to the call.
That is, given <code>f := T.Mv</code>, <code>f</code> is invoked
as <code>f(t, 7)</code> not <code>t.f(7)</code>.
To construct a function that binds the receiver, use a
<a href="#Function_literals">closure</a>.
</p>

<p>
It is legal to derive a function value from a method of an interface type.
The resulting function takes an explicit receiver of that interface type.
</p>

<h3 id="Conversions">Conversions</h3>

<p>
Conversions are expressions of the form <code>T(x)</code>
where <code>T</code> is a type and <code>x</code> is an expression
that can be converted to type <code>T</code>.
</p>

<pre class="ebnf">
Conversion = Type "(" Expression ")" .
</pre>

<p>
If the type starts with an operator it must be parenthesized:
</p>

<pre>
*Point(p)        // same as *(Point(p))
(*Point)(p)      // p is converted to (*Point)
&lt;-chan int(c)    // same as &lt;-(chan int(c))
(&lt;-chan int)(c)  // c is converted to (&lt;-chan int)
</pre>

<p>
In general, a conversion is permitted if
</p>
<ol>
<li>
the value of <code>x</code> would be
<a href="#Assignment_compatibility">assignment compatible</a> with type
<code>T</code> if <code>T</code> were unnamed
</li>
<li>
<code>x</code> is of an unnamed pointer type and type <code>T</code> is another
unnamed pointer type and the previous rule applies to the pointer base types.
</li>
</ol>
<p>
Such a conversion changes the type but not the representation of <code>x</code>.
</p>

<p>
Specific rules apply to conversions where <code>T</code> is a
numeric or string type, or where <code>x</code> is of string type.
These conversions may change the representation of a value and incur a run-time cost.
</p>

<h4>Conversions between integer types</h4>
<p>
If the value is a signed quantity, it is
sign extended to implicit infinite precision; otherwise it is zero
extended.  It is then truncated to fit in the result type's size.
For example, if <code>x := uint16(0x10F0)</code>, then <code>uint32(int8(x)) == 0xFFFFFFF0</code>.
The conversion always yields a valid value; there is no indication of overflow.
</p>

<h4>Conversions involving floating point and complex types</h4>
<ol>
<li>
When converting a floating-point number to an integer, the fraction is discarded
(truncation towards zero).
</li>
<li>
A value of complex type may be converted to a different complex type,
but there is no conversion between complex and any other type.
</li>
<li>
When converting a number to a floating-point or complex type,
the result value is rounded
to the precision specified by the destination type.
For instance, the value of a variable <code>x</code> of type <code>float32</code>
may be stored using additional precision beyond that of an IEEE-754 32-bit number,
but float32(x) represents the result of rounding <code>x</code>'s value to
32-bit precision. Similarly, <code>x + 0.1</code> may use more than 32 bits
of precision, but <code>float32(x + 0.1)</code> does not.
</li>
</ol>

<p>
In all conversions involving floating-point or complex values,
if the result type cannot represent the value the conversion
succeeds but the result value is
implementation-dependent.
</p>

<h4>Conversions to and from a string type</h4>

<ol>
<li>
Converting a signed or unsigned integer value to a string type yields a
string containing the UTF-8 representation of the integer.
Negative values are converted to <code>"\uFFFD"</code>.

<pre>
string('a')           // "a"
string(-1)            // "\ufffd" == "\xef\xbf\xbd "
string(0xf8)          // "\u00f8" == "ø" == "\xc3\xb8"
type MyString string
MyString(0x65e5)      // "\u65e5" == "日" == "\xe6\x97\xa5"
</pre>
</li>

<li>
Converting a value of type <code>[]byte</code> (or
the equivalent <code>[]uint8</code>) to a string type yields a
string whose successive bytes are the elements of the slice.  If
the slice value is <code>nil</code>, the result is the empty string.

<pre>
string([]byte{'h', 'e', 'l', 'l', '\xc3', '\xb8'})  // "hellø"
</pre>
</li>

<li>
Converting a value of type <code>[]int</code> to a string type yields
a string that is the concatenation of the individual integers
converted to strings.  If the slice value is <code>nil</code>, the
result is the empty string.
<pre>
string([]int{0x767d, 0x9d6c, 0x7fd4})  // "\u767d\u9d6c\u7fd4" == "白鵬翔"
</pre>
</li>

<li>
Converting a value of a string type to <code>[]byte</code> (or <code>[]uint8</code>)
yields a slice whose successive elements are the bytes of the string.
If the string is empty, the result is <code>[]byte(nil)</code>.

<pre>
[]byte("hellø")  // []byte{'h', 'e', 'l', 'l', '\xc3', '\xb8'}
</pre>
</li>

<li>
Converting a value of a string type to <code>[]int</code> yields a
slice containing the individual Unicode code points of the string.
If the string is empty, the result is <code>[]int(nil)</code>.
<pre>
[]int(MyString("白鵬翔"))  // []int{0x767d, 0x9d6c, 0x7fd4}
</pre>
</li>
</ol>

<p>
There is no linguistic mechanism to convert between pointers and integers.
The package <a href="#Package_unsafe"><code>unsafe</code></a>
implements this functionality under
restricted circumstances.
</p>

<h3 id="Constant_expressions">Constant expressions</h3>

<p>
Constant expressions may contain only <a href="#Constants">constant</a>
operands and are evaluated at compile-time.
</p>

<p>
Untyped boolean, numeric, and string constants may be used as operands
wherever it is legal to use an operand of boolean, numeric, or string type,
respectively. Except for shift operations, if the operands of a binary operation
are an untyped integer constant and an untyped floating-point constant,
the integer constant is converted to an untyped floating-point constant
(relevant for <code>/</code> and <code>%</code>).
Similarly,
untyped integer or floating-point constants may be used as operands
wherever it is legal to use an operand of complex type;
the integer or floating point constant is converted to a
complex constant with a zero imaginary part.
</p>

<p>
Applying an operator to untyped constants results in an untyped
constant of the same kind (that is, a boolean, integer, floating-point,
complex, or string constant), except for
<a href="#Comparison_operators">comparison operators</a>, which result in
a constant of type <code>bool</code>.
</p>

<p>
Imaginary literals are untyped complex constants (with zero real part)
and may be combined in binary
operations with untyped integer and floating-point constants; the
result is an untyped complex constant.
Complex constants are always constructed from
constant expressions involving imaginary
literals or constants derived from them, or calls of the built-in function
<a href="#Complex_numbers"><code>cmplx</code></a>.
</p>

<pre>
const Σ = 1 - 0.707i
const Δ = Σ + 2.0e-4 - 1/1i
const Φ = iota * 1i
const iΓ = cmplx(0, Γ)
</pre>

<p>
Constant expressions are always evaluated exactly; intermediate values and the
constants themselves may require precision significantly larger than supported
by any predeclared type in the language. The following are legal declarations:
</p>

<pre>
const Huge = 1 &lt;&lt; 100
const Four int8 = Huge &gt;&gt; 98
</pre>

<p>
The values of <i>typed</i> constants must always be accurately representable as values
of the constant type. The following constant expressions are illegal:
</p>

<pre>
uint(-1)       // -1 cannot be represented as a uint
int(3.14)      // 3.14 cannot be represented as an int
int64(Huge)    // 1&lt;&lt;100 cannot be represented as an int64
Four * 300     // 300 cannot be represented as an int8
Four * 100     // 400 cannot be represented as an int8
</pre>

<p>
The mask used by the unary bitwise complement operator <code>^</code> matches
the rule for non-constants: the mask is all 1s for unsigned constants
and -1 for signed and untyped constants.
</p>

<pre>
^1          // untyped integer constant, equal to -2
uint8(^1)   // error, same as uint8(-2), out of range
^uint8(1)   // typed uint8 constant, same as 0xFF ^ uint8(1) = uint8(0xFE)
int8(^1)    // same as int8(-2)
^int8(1)    // same as -1 ^ int8(1) = -2
</pre>

<!---
<p>
<span class="alert">
TODO: perhaps ^ should be disallowed on non-uints instead of assuming twos complement.
Also it may be possible to make typed constants more like variables, at the cost of fewer
overflow etc. errors being caught.
</span>
</p>
--->

<h3 id="Order_of_evaluation">Order of evaluation</h3>

<p>
When evaluating the elements of an assignment or expression,
all function calls, method calls and
communication operations are evaluated in lexical left-to-right
order.
</p>

<p>
For example, in the assignment
</p>
<pre>
y[f()], ok = g(h(), i() + x[j()], &lt;-c), k()
</pre>
<p>
the function calls and communication happen in the order
<code>f()</code>, <code>h()</code>, <code>i()</code>, <code>j()</code>,
<code>&lt;-c</code>, <code>g()</code>, and <code>k()</code>.
However, the order of those events compared to the evaluation
and indexing of <code>x</code> and the evaluation
of <code>y</code> is not specified.
</p>

<p>
Floating-point operations within a single expression are evaluated according to
the associativity of the operators.  Explicit parentheses affect the evaluation
by overriding the default associativity.
In the expression <code>x + (y + z)</code> the addition <code>y + z</code>
is performed before adding <code>x</code>.
</p>

<h2 id="Statements">Statements</h2>

<p>
Statements control execution.
</p>

<pre class="ebnf">
Statement =
	Declaration | LabeledStmt | SimpleStmt |
	GoStmt | ReturnStmt | BreakStmt | ContinueStmt | GotoStmt |
	FallthroughStmt | Block | IfStmt | SwitchStmt | SelectStmt | ForStmt |
	DeferStmt .

SimpleStmt = EmptyStmt | ExpressionStmt | IncDecStmt | Assignment | ShortVarDecl .
</pre>


<h3 id="Empty_statements">Empty statements</h3>

<p>
The empty statement does nothing.
</p>

<pre class="ebnf">
EmptyStmt = .
</pre>


<h3 id="Labeled_statements">Labeled statements</h3>

<p>
A labeled statement may be the target of a <code>goto</code>,
<code>break</code> or <code>continue</code> statement.
</p>

<pre class="ebnf">
LabeledStmt = Label ":" Statement .
Label       = identifier .
</pre>

<pre>
Error: log.Crash("error encountered")
</pre>


<h3 id="Expression_statements">Expression statements</h3>

<p>
Function calls, method calls, and channel operations
can appear in statement context.
</p>


<pre class="ebnf">
ExpressionStmt = Expression .
</pre>

<pre>
f(x+y)
&lt;-ch
</pre>


<h3 id="IncDec_statements">IncDec statements</h3>

<p>
The "++" and "--" statements increment or decrement their operands
by the untyped <a href="#Constants">constant</a> <code>1</code>.
As with an assignment, the operand must be a variable, pointer indirection,
field selector or index expression.
</p>

<pre class="ebnf">
IncDecStmt = Expression ( "++" | "--" ) .
</pre>

<p>
The following <a href="#Assignments">assignment statements</a> are semantically
equivalent:
</p>

<pre class="grammar">
IncDec statement    Assignment
x++                 x += 1
x--                 x -= 1
</pre>

<h3 id="Assignments">Assignments</h3>

<pre class="ebnf">
Assignment = ExpressionList assign_op ExpressionList .

assign_op = [ add_op | mul_op ] "=" .
</pre>

<p>
Each left-hand side operand must be <a href="#Address_operators">addressable</a>,
a map index expression,
or the <a href="#Blank_identifier">blank identifier</a>.
</p>

<pre>
x = 1
*p = f()
a[i] = 23
k = &lt;-ch
</pre>

<p>
An <i>assignment operation</i> <code>x</code> <i>op</i><code>=</code>
<code>y</code> where <i>op</i> is a binary arithmetic operation is equivalent
to <code>x</code> <code>=</code> <code>x</code> <i>op</i>
<code>y</code> but evaluates <code>x</code>
only once.  The <i>op</i><code>=</code> construct is a single token.
In assignment operations, both the left- and right-hand expression lists
must contain exactly one single-valued expression.
</p>

<pre>
a[i] &lt;&lt;= 2
i &amp;^= 1&lt;&lt;n
</pre>

<p>
A tuple assignment assigns the individual elements of a multi-valued
operation to a list of variables.  There are two forms.  In the
first, the right hand operand is a single multi-valued expression
such as a function evaluation or <a href="#Channel_types">channel</a> or
<a href="#Map_types">map</a> operation or a <a href="#Type_assertions">type assertion</a>.
The number of operands on the left
hand side must match the number of values.  For instance, if
<code>f</code> is a function returning two values,
</p>

<pre>
x, y = f()
</pre>

<p>
assigns the first value to <code>x</code> and the second to <code>y</code>.
The <a href="#Blank_identifier">blank identifier</a> provides a
way to ignore values returned by a multi-valued expression:
</p>

<pre>
x, _ = f()  // ignore second value returned by f()
</pre>

<p>
In the second form, the number of operands on the left must equal the number
of expressions on the right, each of which must be single-valued, and the
<i>n</i>th expression on the right is assigned to the <i>n</i>th
operand on the left.
The expressions on the right are evaluated before assigning to
any of the operands on the left, but otherwise the evaluation
order is unspecified beyond <a href="#Order_of_evaluation">the usual rules</a>.
</p>

<pre>
a, b = b, a  // exchange a and b
</pre>

<p>
In assignments, each value must be
<a href="#Assignment_compatibility">assignment compatible</a> with the type of the
operand to which it is assigned. If an untyped <a href="#Constants">constant</a>
is assigned to a variable of interface type, the constant is <a href="#Conversions">converted</a>
to type <code>bool</code>, <code>int</code>, <code>float</code>,
<code>complex</code> or <code>string</code>
respectively, depending on whether the value is a boolean, integer, floating-point,
complex, or string constant.
</p>


<h3 id="If_statements">If statements</h3>

<p>
"If" statements specify the conditional execution of two branches
according to the value of a boolean expression.  If the expression
evaluates to true, the "if" branch is executed, otherwise, if
present, the "else" branch is executed.  A missing condition
is equivalent to <code>true</code>.
</p>

<pre class="ebnf">
IfStmt    = "if" [ SimpleStmt ";" ] [ Expression ] Block [ "else" Statement ] .
</pre>

<pre>
if x > 0 {
	return true;
}
</pre>

<p>
The expression may be preceded by a simple statement, which
executes before the expression is evaluated.
</p>

<pre>
if x := f(); x &lt; y {
	return x
} else if x > z {
	return z
} else {
	return y
}
</pre>


<h3 id="Switch_statements">Switch statements</h3>

<p>
"Switch" statements provide multi-way execution.
An expression or type specifier is compared to the "cases"
inside the "switch" to determine which branch
to execute.
</p>

<pre class="ebnf">
SwitchStmt = ExprSwitchStmt | TypeSwitchStmt .
</pre>

<p>
There are two forms: expression switches and type switches.
In an expression switch, the cases contain expressions that are compared
against the value of the switch expression.
In a type switch, the cases contain types that are compared against the
type of a specially annotated switch expression.
</p>

<h4 id="Expression_switches">Expression switches</h4>

<p>
In an expression switch,
the switch expression is evaluated and
the case expressions, which need not be constants,
are evaluated left-to-right and top-to-bottom; the first one that equals the
switch expression
triggers execution of the statements of the associated case;
the other cases are skipped.
If no case matches and there is a "default" case,
its statements are executed.
There can be at most one default case and it may appear anywhere in the
"switch" statement.
A missing switch expression is equivalent to
the expression <code>true</code>.
</p>

<pre class="ebnf">
ExprSwitchStmt = "switch" [ SimpleStmt ";" ] [ Expression ] "{" { ExprCaseClause } "}" .
ExprCaseClause = ExprSwitchCase ":" { Statement ";" } .
ExprSwitchCase = "case" ExpressionList | "default" .
</pre>

<p>
In a case or default clause,
the last statement only may be a "fallthrough" statement
(§<a href="#Fallthrough_statements">Fallthrough statement</a>) to
indicate that control should flow from the end of this clause to
the first statement of the next clause.
Otherwise control flows to the end of the "switch" statement.
</p>

<p>
The expression may be preceded by a simple statement, which
executes before the expression is evaluated.
</p>

<pre>
switch tag {
default: s3()
case 0, 1, 2, 3: s1()
case 4, 5, 6, 7: s2()
}

switch x := f(); {  // missing switch expression means "true"
case x &lt; 0: return -x
default: return x
}

switch {
case x &lt; y: f1()
case x &lt; z: f2()
case x == 4: f3()
}
</pre>

<h4 id="Type_switches">Type switches</h4>

<p>
A type switch compares types rather than values. It is otherwise similar
to an expression switch. It is marked by a special switch expression that
has the form of a <a href="#Type_assertions">type assertion</a>
using the reserved word <code>type</code> rather than an actual type.
Cases then match literal types against the dynamic type of the expression
in the type assertion.
</p>

<pre class="ebnf">
TypeSwitchStmt  = "switch" [ SimpleStmt ";" ] TypeSwitchGuard "{" { TypeCaseClause } "}" .
TypeSwitchGuard = [ identifier ":=" ] PrimaryExpr "." "(" "type" ")" .
TypeCaseClause  = TypeSwitchCase ":" { Statement ";" } .
TypeSwitchCase  = "case" TypeList | "default" .
TypeList        = Type { "," Type } .
</pre>

<p>
The TypeSwitchGuard may include a
<a href="#Short_variable_declarations">short variable declaration</a>.
When that form is used, the variable is declared in each clause.
In clauses with a case listing exactly one type, the variable
has that type; otherwise, the variable has the type of the expression
in the TypeSwitchGuard.
</p>

<p>
The type in a case may be <code>nil</code>
(§<a href="#Predeclared_identifiers">Predeclared identifiers</a>);
that case is used when the expression in the TypeSwitchGuard
is a <code>nil</code> interface value.
</p>

<p>
Given an expression <code>x</code> of type <code>interface{}</code>,
the following type switch:
</p>

<pre>
switch i := x.(type) {
case nil:
	printString("x is nil")
case int:
	printInt(i)  // i is an int
case float:
	printFloat(i)  // i is a float
case func(int) float:
	printFunction(i)  // i is a function
case bool, string:
	printString("type is bool or string")  // i is an interface{}
default:
	printString("don't know the type")
}
</pre>

<p>
could be rewritten:
</p>

<pre>
v := x  // x is evaluated exactly once
if v == nil {
	printString("x is nil")
} else if i, is_int := v.(int); is_int {
	printInt(i)  // i is an int
} else if i, is_float := v.(float); is_float {
	printFloat(i)  // i is a float
} else if i, is_func := v.(func(int) float); is_func {
	printFunction(i)  // i is a function
} else {
	i1, is_bool := v.(bool)
	i2, is_string := v.(string)
	if is_bool || is_string {
		i := v
		printString("type is bool or string")  // i is an interface{}
	} else {
		i := v
		printString("don't know the type")  // i is an interface{}
	}
}
</pre>

<p>
The type switch guard may be preceded by a simple statement, which
executes before the guard is evaluated.
</p>

<p>
The "fallthrough" statement is not permitted in a type switch.
</p>

<h3 id="For_statements">For statements</h3>

<p>
A "for" statement specifies repeated execution of a block. The iteration is
controlled by a condition, a "for" clause, or a "range" clause.
</p>

<pre class="ebnf">
ForStmt = "for" [ Condition | ForClause | RangeClause ] Block .
Condition = Expression .
</pre>

<p>
In its simplest form, a "for" statement specifies the repeated execution of
a block as long as a boolean condition evaluates to true.
The condition is evaluated before each iteration.
If the condition is absent, it is equivalent to <code>true</code>.
</p>

<pre>
for a &lt; b {
	a *= 2
}
</pre>

<p>
A "for" statement with a ForClause is also controlled by its condition, but
additionally it may specify an <i>init</i>
and a <i>post</i> statement, such as an assignment,
an increment or decrement statement. The init statement may be a
<a href="#Short_variable_declarations">short variable declaration</a>, but the post statement must not.
</p>

<pre class="ebnf">
ForClause = [ InitStmt ] ";" [ Condition ] ";" [ PostStmt ] .
InitStmt = SimpleStmt .
PostStmt = SimpleStmt .
</pre>

<pre>
for i := 0; i &lt; 10; i++ {
	f(i)
}
</pre>

<p>
If non-empty, the init statement is executed once before evaluating the
condition for the first iteration;
the post statement is executed after each execution of the block (and
only if the block was executed).
Any element of the ForClause may be empty but the
<a href="#Semicolons">semicolons</a> are
required unless there is only a condition.
If the condition is absent, it is equivalent to <code>true</code>.
</p>

<pre>
for cond { S() }    is the same as    for ; cond ; { S() }
for      { S() }    is the same as    for true     { S() }
</pre>

<p>
A "for" statement with a "range" clause
iterates through all entries of an array, slice, string or map,
or values received on a channel.
For each entry it first assigns the current index or key to an iteration
variable - or the current (index, element) or (key, value) pair to a pair
of iteration variables - and then executes the block.
</p>

<pre class="ebnf">
RangeClause = ExpressionList ( "=" | ":=" ) "range" Expression .
</pre>

<p>
The type of the right-hand expression in the "range" clause must be an
array, slice, string or map, or a pointer to an array;
or it may be a channel.
Except for channels,
the identifier list must contain one or two expressions
(as in assignments, these must be a
variable, pointer indirection, field selector, or index expression)
denoting the
iteration variables. On each iteration,
the first variable is set to the string, array or slice index or
map key, and the second variable, if present, is set to the corresponding
string or array element or map value.
The types of the array or slice index (always <code>int</code>)
and element, or of the map key and value respectively,
must be <a href="#Assignment_compatibility">assignment compatible</a> with
the type of the iteration variables.  The expression on the right hand
side is evaluated once before beginning the loop.  At each iteration
of the loop, the values produced by the range clause are assigned to
the left hand side as in an <a href="#Assignments">assignment
statement</a>.  Function calls on the left hand side will be evaluated
exactly once per iteration.
</p>
<p>
For a value of a string type, the "range" clause iterates over the Unicode code points
in the string.  On successive iterations, the index variable will be the
index of the first byte of successive UTF-8-encoded code points in the string, and
the second variable, of type <code>int</code>, will be the value of
the corresponding code point.  If the iteration encounters an invalid
UTF-8 sequence, the second variable will be <code>0xFFFD</code>,
the Unicode replacement character, and the next iteration will advance
a single byte in the string.
</p>
<p>
For channels, the identifier list must contain one identifier.
The iteration receives values sent on the channel until the channel is closed;
it does not process the zero value sent before the channel is closed.
</p>
<p>
The iteration variables may be declared by the "range" clause (":="), in which
case their scope ends at the end of the "for" statement (§<a href="#Declarations_and">Declarations and</a>
scope rules). In this case their types are set to
<code>int</code> and the array element type, or the map key and value types, respectively.
If the iteration variables are declared outside the "for" statement,
after execution their values will be those of the last iteration.
</p>

<pre>
var a [10]string
m := map[string]int{"mon":0, "tue":1, "wed":2, "thu":3, "fri":4, "sat":5, "sun":6}

for i, s := range a {
	// type of i is int
	// type of s is string
	// s == a[i]
	g(i, s)
}

var key string
var val interface {}  // value type of m is assignment compatible with val
for key, val = range m {
	h(key, val)
}
// key == last map key encountered in iteration
// val == map[key]
</pre>

<p>
If map entries that have not yet been processed are deleted during iteration,
they will not be processed. If map entries are inserted during iteration, the
behavior is implementation-dependent, but each entry will be processed at most once.
</p>

<h3 id="Go_statements">Go statements</h3>

<p>
A "go" statement starts the execution of a function or method call
as an independent concurrent thread of control, or <i>goroutine</i>,
within the same address space.
</p>

<pre class="ebnf">
GoStmt = "go" Expression .
</pre>

<p>
The expression must be a call, and
unlike with a regular call, program execution does not wait
for the invoked function to complete.
</p>

<pre>
go Server()
go func(ch chan&lt;- bool) { for { sleep(10); ch &lt;- true; }} (c)
</pre>


<h3 id="Select_statements">Select statements</h3>

<p>
A "select" statement chooses which of a set of possible communications
will proceed.  It looks similar to a "switch" statement but with the
cases all referring to communication operations.
</p>

<pre class="ebnf">
SelectStmt = "select" "{" { CommClause } "}" .
CommClause = CommCase ":" { Statement ";" } .
CommCase = "case" ( SendExpr | RecvExpr) | "default" .
SendExpr =  Expression "&lt;-" Expression .
RecvExpr =  [ Expression ( "=" | ":=" ) ] "&lt;-" Expression .
</pre>

<p>
For all the send and receive expressions in the "select"
statement, the channel expressions are evaluated, along with
any expressions that appear on the right hand side of send expressions,
in top-to-bottom order.
If any of the resulting operations can proceed, one is
chosen and the corresponding communication and statements are
evaluated.  Otherwise, if there is a default case, that executes;
if not, the statement blocks until one of the communications can
complete.  The channels and send expressions are not re-evaluated.
A channel pointer may be <code>nil</code>,
which is equivalent to that case not
being present in the select statement
except, if a send, its expression is still evaluated.
</p>
<p>
Since all the channels and send expressions are evaluated, any side
effects in that evaluation will occur for all the communications
in the "select" statement.
</p>
<p>
If multiple cases can proceed, a uniform fair choice is made to decide
which single communication will execute.
<p>
The receive case may declare a new variable using a
<a href="#Short_variable_declarations">short variable declaration</a>.
</p>

<pre>
var c, c1, c2 chan int
var i1, i2 int
select {
case i1 = &lt;-c1:
	print("received ", i1, " from c1\n")
case c2 &lt;- i2:
	print("sent ", i2, " to c2\n")
default:
	print("no communication\n")
}

for {  // send random sequence of bits to c
	select {
	case c &lt;- 0:  // note: no statement, no fallthrough, no folding of cases
	case c &lt;- 1:
	}
}
</pre>


<h3 id="Return_statements">Return statements</h3>

<p>
A "return" statement terminates execution of the containing function
and optionally provides a result value or values to the caller.
</p>

<pre class="ebnf">
ReturnStmt = "return" [ ExpressionList ] .
</pre>

<p>
In a function without a result type, a "return" statement must not
specify any result values.
</p>
<pre>
func no_result() {
	return
}
</pre>

<p>
There are three ways to return values from a function with a result
type:
</p>

<ol>
	<li>The return value or values may be explicitly listed
		in the "return" statement. Each expression must be single-valued
		and <a href="#Assignment_compatibility">assignment compatible</a>
		with the type of the corresponding element of the function's
		result type.
<pre>
func simple_f() int {
	return 2
}

func complex_f1() (re float, im float) {
	return -7.0, -4.0
}
</pre>
	</li>
	<li>The expression list in the "return" statement may be a single
		call to a multi-valued function. The effect is as if each value
		returned from that function were assigned to a temporary
		variable with the type of the respective value, followed by a
		"return" statement listing these variables, at which point the
		rules of the previous case apply.
<pre>
func complex_f2() (re float, im float) {
	return complex_f1()
}
</pre>
	</li>
	<li>The expression list may be empty if the functions's result
		type specifies names for its result parameters (§<a href="#Function_Types">Function Types</a>).
		The result parameters act as ordinary local variables that are
		initialized to the zero values for their type (§<a href="#The_zero_value">The zero value</a>)
		and the function may assign values to them as necessary.
		The "return" statement returns the values of these variables.
<pre>
func complex_f3() (re float, im float) {
	re = 7.0
	im = 4.0
	return
}
</pre>
	</li>
</ol>

<!---
<p>
<span class="alert">
TODO: Define when return is required.<br />
TODO: Language about result parameters needs to go into a section on
      function/method invocation<br />
</span>
</p>
--->

<h3 id="Break_statements">Break statements</h3>

<p>
A "break" statement terminates execution of the innermost
"for", "switch" or "select" statement.
</p>

<pre class="ebnf">
BreakStmt = "break" [ Label ] .
</pre>

<p>
If there is a label, it must be that of an enclosing
"for", "switch" or "select" statement, and that is the one whose execution
terminates
(§<a href="#For_statements">For statements</a>, §<a href="#Switch_statements">Switch statements</a>, §<a href="#Select_statements">Select statements</a>).
</p>

<pre>
L: for i &lt; n {
	switch i {
		case 5: break L
	}
}
</pre>

<h3 id="Continue_statements">Continue statements</h3>

<p>
A "continue" statement begins the next iteration of the
innermost "for" loop at its post statement (§<a href="#For_statements">For statements</a>).
</p>

<pre class="ebnf">
ContinueStmt = "continue" [ Label ] .
</pre>

<p>
If there is a label, it must be that of an enclosing
"for" statement, and that is the one whose execution
advances
(§<a href="#For_statements">For statements</a>).
</p>

<h3 id="Goto_statements">Goto statements</h3>

<p>
A "goto" statement transfers control to the statement with the corresponding label.
</p>

<pre class="ebnf">
GotoStmt = "goto" Label .
</pre>

<pre>
goto Error
</pre>

<p>
Executing the "goto" statement must not cause any variables to come into
scope that were not already in scope at the point of the goto.  For
instance, this example:
</p>

<pre>
goto L  // BAD
v := 3
L:
</pre>

<p>
is erroneous because the jump to label <code>L</code> skips
the creation of <code>v</code>.
<!---
(<span class="alert">TODO: Eliminate in favor of used and not set errors?</span>)
--->
</p>

<h3 id="Fallthrough_statements">Fallthrough statements</h3>

<p>
A "fallthrough" statement transfers control to the first statement of the
next case clause in a expression "switch" statement (§<a href="#Expression_switches">Expression switches</a>). It may
be used only as the final non-empty statement in a case or default clause in an
expression "switch" statement.
</p>

<pre class="ebnf">
FallthroughStmt = "fallthrough" .
</pre>


<h3 id="Defer_statements">Defer statements</h3>

<p>
A "defer" statement invokes a function whose execution is deferred to the moment
the surrounding function returns.
</p>

<pre class="ebnf">
DeferStmt = "defer" Expression .
</pre>

<p>
The expression must be a function or method call.
Each time the "defer" statement
executes, the parameters to the function call are evaluated and saved anew but the
function is not invoked.
Deferred function calls are executed in LIFO order
immediately before the surrounding function returns,
after the return values, if any, have been evaluated, but before they
are returned to the caller. For instance, if the deferred function is
a <a href="#Function_literals">function literal</a> and the surrounding
function has <a href="#Function_types">named result parameters</a> that
are in scope within the literal, the deferred function may access and modify
the result parameters before they are returned.
</p>

<pre>
lock(l)
defer unlock(l)  // unlocking happens before surrounding function returns

// prints 3 2 1 0 before surrounding function returns
for i := 0; i &lt;= 3; i++ {
	defer fmt.Print(i)
}

// f returns 1
func f() (result int) {
	defer func() {
		result++
	}()
	return 0
}
</pre>

<h2 id="Built-in_functions">Built-in functions</h2>

<p>
Built-in functions are
<a href="#Predeclared_identifiers">predeclared</a>.
They are called like any other function but some of them
accept a type instead of an expression as the first argument.
</p>

<p>
The built-in functions do not have standard Go types,
so they can only appear in <a href="#Calls">call expressions</a>;
they cannot be used as function values.
</p>

<pre class="ebnf">
BuiltinCall = identifier "(" [ BuiltinArgs ] ")" .
BuiltinArgs = Type [ "," ExpressionList ] | ExpressionList .
</pre>

<h3 id="Close_and_closed">Close and closed</h3>

<p>
For a channel <code>c</code>, the predefined function <code>close(c)</code>
marks the channel as unable to accept more
values through a send operation.  After any previously
sent values have been received, receive operations will return
the zero value for the channel's type.  After at least one such zero value has been
received, <code>closed(c)</code> returns true.
</p>

<h3 id="Length_and_capacity">Length and capacity</h3>

<p>
The built-in functions <code>len</code> and <code>cap</code> take arguments
of various types and return a result of type <code>int</code>.
The implementation guarantees that the result always fits into an <code>int</code>.
</p>

<pre class="grammar">
Call      Argument type        Result

len(s)    string type          string length in bytes
          [n]T, *[n]T          array length (== constant n)
          []T                  slice length
          map[K]T              map length (number of defined keys)
          chan T               number of elements queued in channel buffer

cap(s)    [n]T, *[n]T          array length (== constant n)
          []T                  slice capacity
          chan T               channel buffer capacity
</pre>

<p>
The capacity of a slice is the number of elements for which there is
space allocated in the underlying array.
At any time the following relationship holds:
</p>

<pre>
0 <= len(s) <= cap(s)
</pre>


<h3 id="Allocation">Allocation</h3>

<p>
The built-in function <code>new</code> takes a type <code>T</code> and
returns a value of type <code>*T</code>.
The memory is initialized as described in the section on initial values
(§<a href="#The_zero_value">The zero value</a>).
</p>

<pre class="grammar">
new(T)
</pre>

<p>
For instance
</p>

<pre>
type S struct { a int; b float }
new(S)
</pre>

<p>
dynamically allocates memory for a variable of type <code>S</code>,
initializes it (<code>a=0</code>, <code>b=0.0</code>),
and returns a value of type <code>*S</code> containing the address
of the memory.
</p>

<h3 id="Making_slices_maps_and_channels">Making slices, maps and channels</h3>

<p>
Slices, maps and channels are reference types that do not require the
extra indirection of an allocation with <code>new</code>.
The built-in function <code>make</code> takes a type <code>T</code>,
which must be a slice, map or channel type,
optionally followed by a type-specific list of expressions.
It returns a value of type <code>T</code> (not <code>*T</code>).
The memory is initialized as described in the section on initial values
(§<a href="#The_zero_value">The zero value</a>).
</p>

<pre class="grammar">
Call             Type T     Result

make(T, n)       slice      slice of type T with length n and capacity n
make(T, n, m)    slice      slice of type T with length n and capacity m

make(T)          map        map of type T
make(T, n)       map        map of type T with initial space for n elements

make(T)          channel    synchronous channel of type T
make(T, n)       channel    asynchronous channel of type T, buffer size n
</pre>


<p>
The arguments <code>n</code> and <code>m</code> must be of integer type.
A <a href="#Run_time_panics">run-time panic</a> occurs if <code>n</code>
is negative or larger than <code>m</code>, or if <code>n</code> or
<code>m</code> cannot be represented by an <code>int</code>.
</p>

<pre>
s := make([]int, 10, 100)        // slice with len(s) == 10, cap(s) == 100
s := make([]int, 10)             // slice with len(s) == cap(s) == 10
c := make(chan int, 10)          // channel with a buffer size of 10
m := make(map[string] int, 100)  // map with initial space for 100 elements
</pre>


<h3 id="Copying_slices">Copying slices</h3>

<p>
The built-in function <code>copy</code> copies slice elements from
a source <code>src</code> to a destination <code>dst</code> and returns the
number of elements copied. Source and destination may overlap.
Both arguments must have identical element type <code>T</code> and must be
<a href="#Assignment_compatibility">assignment compatible</a> to a slice
of type <code>[]T</code>. The number of arguments copied is the minimum of
<code>len(src)</code> and <code>len(dst)</code>.
</p>

<pre class="grammar">
copy(dst, src []T) int
</pre>

<p>
Examples:
</p>

<pre>
var a = [...]int{0, 1, 2, 3, 4, 5, 6, 7}
var s = make([]int, 6)
n1 := copy(s, a[0:])  // n1 == 6, s == []int{0, 1, 2, 3, 4, 5}
n2 := copy(s, s[2:])  // n2 == 4, s == []int{2, 3, 4, 5, 4, 5}
</pre>

<h3 id="Complex_numbers">Assembling and disassembling complex numbers</h3>

<p>
Three functions assemble and disassemble complex numbers.
The built-in function <code>cmplx</code> constructs a complex
value from a floating-point real and imaginary part, while
<code>real</code> and <code>imag</code>
extract the real and imaginary parts of a complex value.
</p>

<pre class="grammar">
cmplx(realPart, imaginaryPart floatT) complexT
real(complexT) floatT
imag(complexT) floatT
</pre>

<p>
The type of the arguments and return value correspond.
For <code>cmplx</code>, the two arguments must be of the same
floating-point type and the return type is the complex type
with the corresponding floating-point constituents:
<code>complex</code> for <code>float</code>,
<code>complex64</code> for <code>float32</code>,
<code>complex128</code> for <code>float64</code>.
The <code>real</code> and <code>imag</code> functions
together form the inverse, so for a complex value <code>z</code>,
<code>z</code> <code>==</code> <code>cmplx(real(z),</code> <code>imag(z))</code>.
</p>

<p>
If the operands of these functions are all constants, the return
value is a constant.
</p>

<pre>
var a = cmplx(2, -2)  // has type complex
var b = cmplx(1.0, -1.4)  // has type complex
x := float32(math.Cos(math.Pi/2))
var c64 = cmplx(5, -x)  // has type complex64
var im = imag(b)  // has type float
var rl = real(c64)  // type float32
</pre>

<h3 id="Handling_panics">Handling panics</h3>

<p> Two built-in functions, <code>panic</code> and <code>recover</code>,
assist in reporting and handling <a href="#Run_time_panics">run-time panics</a>
and program-defined error conditions. 
</p>

<pre class="grammar">
func panic(interface{})
func recover() interface{}
</pre>

<p>
<font color=red>TODO: Most of this text could move to the respective
comments in <code>runtime.go</code> once the functions are implemented.
They are here, at least for now, for reference and discussion.
</font>
</p>

<p>
When a function <code>F</code> calls <code>panic</code>, normal
execution of <code>F</code> stops immediately.  Any functions whose
execution was <a href="#Defer_statements">deferred</a> by the
invocation of <code>F</code> are run in the usual way, and then
<code>F</code> returns to its caller.  To the caller, <code>F</code>
then behaves like a call to <code>panic</code>, terminating its own
execution and running deferred functions.  This continues until all
functions in the goroutine have ceased execution, in reverse order.
At that point, the program is
terminated and the error condition is reported, including the value of
the argument to <code>panic</code>.  This termination sequence is
called <i>panicking</i>.
</p>

<p>
The <code>recover</code> function allows a program to manage behavior
of a panicking goroutine.  Executing a <code>recover</code> call
inside a deferred function (but not any function called by it) stops
the panicking sequence by restoring normal execution, and retrieves
the error value passed to the call of <code>panic</code>.  If
<code>recover</code> is called outside the deferred function it will
not stop a panicking sequence.  In this case, and when the goroutine
is not panicking, <code>recover</code> returns <code>nil</code>.
</p>

<p>
If the function defined here,
</p>

<pre>
func f(hideErrors bool) {
	defer func() {
		if x := recover(); x != nil {
			println("panicking with value", x)
			if !hideErrors {
				panic(x)  // go back to panicking
			}
		}
		println("function returns normally") // executes only when hideErrors==true
	}()
	println("before")
	p()
	println("after")	// never executes
}

func p() {
	panic(3)
}
</pre>

<p>
is called with <code>hideErrors=true</code>, it prints
</p>

<pre>
before
panicking with value 3
function returns normally
</pre>

<p>
and resumes normal execution in the function that called <code>f</code>. Otherwise, it prints
</p>

<pre>
before
panicking with value 3
</pre>

<p>
and, absent further <code>recover</code> calls, terminates the program.
</p>

<p>
Since deferred functions run before assigning the return values to the caller
of the deferring function, a deferred invocation of a function literal may modify the
invoking function's return values in the event of a panic. This permits a function to protect its
caller from panics that occur in functions it calls.
</p>

<pre>
func IsPrintable(s string) (ok bool) {
	ok = true
	defer func() {
		if recover() != nil {
			println("input is not printable")
			ok = false
		}
		// Panicking has stopped; execution will resume normally in caller.
		// The return value will be true normally, false if a panic occurred.
	}
	panicIfNotPrintable(s)	// will panic if validations fails.
}
</pre>

<!---
<p>
A deferred function that calls <code>recover</code> will see the
argument passed to <code>panic</code>.  However, functions called
<i>from</i> the deferred function run normally, without behaving as
though they are panicking.  This allows deferred code to run normally
in case recovery is necessary and guarantees that functions that manage
their own panics will not fail incorrectly.  The function
</p>

<pre>
func g() {
	s := ReadString()
	defer func() {
		if IsPrintable(s) {
			println("finished processing", s)
		} else {
			println("finished processing unprintable string")
		}
	}()
	Analyze(s)
}
</pre>

<p>
will not cause <code>IsPrintable</code> to print <code>"input is not printable"</code>
due to a <code>panic</code> triggered by the call to <code>Analyze</code>.
</p>
-->

<h3 id="Bootstrapping">Bootstrapping</h3>

<p>
Current implementations provide several built-in functions useful during
bootstrapping. These functions are documented for completeness but are not
guaranteed to stay in the language. They do not return a result.
</p>

<pre class="grammar">
Function   Behavior

print      prints all arguments; formatting of arguments is implementation-specific
println    like print but prints spaces between arguments and a newline at the end
</pre>


<h2 id="Packages">Packages</h2>

<p>
Go programs are constructed by linking together <i>packages</i>.
A package in turn is constructed from one or more source files
that together declare constants, types, variables and functions
belonging to the package and which are accessible in all files
of the same package. Those elements may be
<a href="#Exported_identifiers">exported</a> and used in another package.
</p>

<h3 id="Source_file_organization">Source file organization</h3>

<p>
Each source file consists of a package clause defining the package
to which it belongs, followed by a possibly empty set of import
declarations that declare packages whose contents it wishes to use,
followed by a possibly empty set of declarations of functions,
types, variables, and constants.
</p>

<pre class="ebnf">
SourceFile       = PackageClause ";" { ImportDecl ";" } { TopLevelDecl ";" } .
</pre>

<h3 id="Package_clause">Package clause</h3>

<p>
A package clause begins each source file and defines the package
to which the file belongs.
</p>

<pre class="ebnf">
PackageClause  = "package" PackageName .
PackageName    = identifier .
</pre>

<p>
The PackageName must not be the <a href="#Blank_identifier">blank identifier</a>.
</p>

<pre>
package math
</pre>

<p>
A set of files sharing the same PackageName form the implementation of a package.
An implementation may require that all source files for a package inhabit the same directory.
</p>

<h3 id="Import_declarations">Import declarations</h3>

<p>
An import declaration states that the source file containing the
declaration uses identifiers
<a href="#Exported_identifiers">exported</a> by the <i>imported</i>
package and enables access to them.  The import names an
identifier (PackageName) to be used for access and an ImportPath
that specifies the package to be imported.
</p>

<pre class="ebnf">
ImportDecl       = "import" ( ImportSpec | "(" { ImportSpec ";" } ")" ) .
ImportSpec       = [ "." | PackageName ] ImportPath .
ImportPath       = string_lit .
</pre>

<p>
The PackageName is used in <a href="#Qualified_identifiers">qualified identifiers</a>
to access the exported identifiers of the package within the importing source file.
It is declared in the <a href="#Blocks">file block</a>.
If the PackageName is omitted, it defaults to the identifier specified in the
<a href="#Package_clauses">package clause</a> of the imported package.
If an explicit period (<code>.</code>) appears instead of a name, all the
package's exported identifiers will be declared in the current file's
file block and can be accessed without a qualifier.
</p>

<p>
The interpretation of the ImportPath is implementation-dependent but
it is typically a substring of the full file name of the compiled
package and may be relative to a repository of installed packages.
</p>

<p>
Assume we have compiled a package containing the package clause
<code>package math</code>, which exports function <code>Sin</code>, and
installed the compiled package in the file identified by
<code>"lib/math"</code>.
This table illustrates how <code>Sin</code> may be accessed in files
that import the package after the
various types of import declaration.
</p>

<pre class="grammar">
Import declaration          Local name of Sin

import   "lib/math"         math.Sin
import M "lib/math"         M.Sin
import . "lib/math"         Sin
</pre>

<p>
An import declaration declares a dependency relation between
the importing and imported package.
It is illegal for a package to import itself or to import a package without
referring to any of its exported identifiers. To import a package solely for
its side-effects (initialization), use the <a href="#Blank_identifier">blank</a>
identifier as explicit package name:
</p>

<pre>
import _ "lib/math"
</pre>


<h3 id="An_example_package">An example package</h3>

<p>
Here is a complete Go package that implements a concurrent prime sieve.
</p>

<pre>
package main

import "fmt"

// Send the sequence 2, 3, 4, ... to channel 'ch'.
func generate(ch chan&lt;- int) {
	for i := 2; ; i++ {
		ch &lt;- i  // Send 'i' to channel 'ch'.
	}
}

// Copy the values from channel 'src' to channel 'dst',
// removing those divisible by 'prime'.
func filter(src &lt;-chan int, dst chan&lt;- int, prime int) {
	for i := range src {	// Loop over values received from 'src'.
		if i%prime != 0 {
			dst &lt;- i  // Send 'i' to channel 'dst'.
		}
	}
}

// The prime sieve: Daisy-chain filter processes together.
func sieve() {
	ch := make(chan int)  // Create a new channel.
	go generate(ch)       // Start generate() as a subprocess.
	for {
		prime := &lt;-ch
		fmt.Print(prime, "\n")
		ch1 := make(chan int)
		go filter(ch, ch1, prime)
		ch = ch1
	}
}

func main() {
	sieve()
}
</pre>

<h2 id="Program_initialization_and_execution">Program initialization and execution</h2>

<h3 id="The_zero_value">The zero value</h3>
<p>
When memory is allocated to store a value, either through a declaration
or <code>make()</code> or <code>new()</code> call,
and no explicit initialization is provided, the memory is
given a default initialization.  Each element of such a value is
set to the <i>zero value</i> for its type: <code>false</code> for booleans,
<code>0</code> for integers, <code>0.0</code> for floats, <code>""</code>
for strings, and <code>nil</code> for pointers, functions, interfaces, slices, channels, and maps.
This initialization is done recursively, so for instance each element of an
array of structs will have its fields zeroed if no value is specified.
</p>
<p>
These two simple declarations are equivalent:
</p>

<pre>
var i int
var i int = 0
</pre>

<p>
After
</p>

<pre>
type T struct { i int; f float; next *T }
t := new(T)
</pre>

<p>
the following holds:
</p>

<pre>
t.i == 0
t.f == 0.0
t.next == nil
</pre>

<p>
The same would also be true after
</p>

<pre>
var t T
</pre>

<h3 id="Program_execution">Program execution</h3>
<p>
A package with no imports is initialized by assigning initial values to
all its package-level variables
and then calling any
package-level function with the name and signature of
</p>
<pre>
func init()
</pre>
<p>
defined in its source.
A package may contain multiple
<code>init()</code> functions, even
within a single source file; they execute
in unspecified order.
</p>
<p>
Within a package, package-level variables are initialized,
and constant values are determined, in
data-dependent order: if the initializer of <code>A</code>
depends on the value of <code>B</code>, <code>A</code>
will be set after <code>B</code>.
It is an error if such dependencies form a cycle.
Dependency analysis is done lexically: <code>A</code>
depends on <code>B</code> if the value of <code>A</code>
contains a mention of <code>B</code>, contains a value
whose initializer
mentions <code>B</code>, or mentions a function that
mentions <code>B</code>, recursively.
If two items are not interdependent, they will be initialized
in the order they appear in the source.
Since the dependency analysis is done per package, it can produce
unspecified results  if <code>A</code>'s initializer calls a function defined
in another package that refers to <code>B</code>.
</p>
<p>
Initialization code may contain "go" statements, but the functions
they invoke do not begin execution until initialization of the entire
program is complete. Therefore, all initialization code is run in a single
goroutine.
</p>
<p>
An <code>init()</code> function cannot be referred to from anywhere
in a program. In particular, <code>init()</code> cannot be called explicitly,
nor can a pointer to <code>init</code> be assigned to a function variable.
</p>
<p>
If a package has imports, the imported packages are initialized
before initializing the package itself. If multiple packages import
a package <code>P</code>, <code>P</code> will be initialized only once.
</p>
<p>
The importing of packages, by construction, guarantees that there can
be no cyclic dependencies in initialization.
</p>
<p>
A complete program, possibly created by linking multiple packages,
must have one package called <code>main</code>, with a function
</p>

<pre>
func main() { ... }
</pre>

<p>
defined.
The function <code>main.main()</code> takes no arguments and returns no value.
</p>
<p>
Program execution begins by initializing the <code>main</code> package and then
invoking <code>main.main()</code>.
</p>
<p>
When <code>main.main()</code> returns, the program exits.  It does not wait for
other (non-<code>main</code>) goroutines to complete.
</p>
<p>
Implementation restriction: The compiler assumes package <code>main</code>
is not imported by any other package.
</p>

<h2 id="Run_time_panics">Run-time panics</h2>

<p>
Execution errors such as attempting to index an array out
of bounds trigger a <i>run-time panic</i> equivalent to a call of
the built-in function <a href="#Handling_panics"><code>panic</code></a>
with a value of the implementation-defined interface type <code>runtime.Error</code>.
That type defines at least the method
<code>String() string</code>.  The exact error values that
represent distinct run-time error conditions are unspecified,
at least for now.
</p>

<pre>
package runtime

type Error interface {
	String() string
	// and perhaps others
}
</pre>

<h2 id="System_considerations">System considerations</h2>

<h3 id="Package_unsafe">Package <code>unsafe</code></h3>

<p>
The built-in package <code>unsafe</code>, known to the compiler,
provides facilities for low-level programming including operations
that violate the type system. A package using <code>unsafe</code>
must be vetted manually for type safety.  The package provides the
following interface:
</p>

<pre class="grammar">
package unsafe

type ArbitraryType int  // shorthand for an arbitrary Go type; it is not a real type
type Pointer *ArbitraryType

func Alignof(variable ArbitraryType) int
func Offsetof(selector ArbitraryType) int
func Sizeof(variable ArbitraryType) int

func Reflect(val interface {}) (typ runtime.Type, addr uintptr)
func Typeof(val interface {}) reflect.Type
func Unreflect(typ runtime.Type, addr uintptr) interface{}
</pre>

<p>
Any pointer or value of type <code>uintptr</code> can be converted into
a <code>Pointer</code> and vice versa.
</p>
<p>
The function <code>Sizeof</code> takes an expression denoting a
variable of any type and returns the size of the variable in bytes.
</p>
<p>
The function <code>Offsetof</code> takes a selector (§<a href="#Selectors">Selectors</a>) denoting a struct
field of any type and returns the field offset in bytes relative to the
struct's address.
For a struct <code>s</code> with field <code>f</code>:
</p>

<pre>
uintptr(unsafe.Pointer(&amp;s)) + uintptr(unsafe.Offsetof(s.f)) == uintptr(unsafe.Pointer(&amp;s.f))
</pre>

<p>
Computer architectures may require memory addresses to be <i>aligned</i>;
that is, for addresses of a variable to be a multiple of a factor,
the variable's type's <i>alignment</i>.  The function <code>Alignof</code>
takes an expression denoting a variable of any type and returns the
alignment of the (type of the) variable in bytes.  For a variable
<code>x</code>:
</p>

<pre>
uintptr(unsafe.Pointer(&amp;x)) % uintptr(unsafe.Alignof(x)) == 0
</pre>

<p>
Calls to <code>Alignof</code>, <code>Offsetof</code>, and
<code>Sizeof</code> are compile-time constant expressions of type <code>int</code>.
</p>
<p>
The functions <code>unsafe.Typeof</code>,
<code>unsafe.Reflect</code>,
and <code>unsafe.Unreflect</code> allow access at run time to the dynamic
types and values stored in interfaces.
<code>Typeof</code> returns a representation of
<code>val</code>'s
dynamic type as a <code>runtime.Type</code>.
<code>Reflect</code> allocates a copy of
<code>val</code>'s dynamic
value and returns both the type and the address of the copy.
<code>Unreflect</code> inverts <code>Reflect</code>,
creating an
interface value from a type and address.
The <a href="/pkg/reflect/"><code>reflect</code> package</a> built on these primitives
provides a safe, more convenient way to inspect interface values.
</p>


<h3 id="Size_and_alignment_guarantees">Size and alignment guarantees</h3>

<p>
For the numeric types (§<a href="#Numeric_types">Numeric types</a>), the following sizes are guaranteed:
</p>

<pre class="grammar">
type                      size in bytes

byte, uint8, int8         1
uint16, int16             2
uint32, int32, float32    4
uint64, int64, float64    8
</pre>

<p>
The following minimal alignment properties are guaranteed:
</p>
<ol>
<li>For a variable <code>x</code> of any type: <code>1 <= unsafe.Alignof(x) <= unsafe.Maxalign</code>.
</li>

<li>For a variable <code>x</code> of numeric type: <code>unsafe.Alignof(x)</code> is the smaller
   of <code>unsafe.Sizeof(x)</code> and <code>unsafe.Maxalign</code>, but at least 1.
</li>

<li>For a variable <code>x</code> of struct type: <code>unsafe.Alignof(x)</code> is the largest of
   all the values <code>unsafe.Alignof(x.f)</code> for each field <code>f</code> of x, but at least 1.
</li>

<li>For a variable <code>x</code> of array type: <code>unsafe.Alignof(x)</code> is the same as
   <code>unsafe.Alignof(x[0])</code>, but at least 1.
</li>
</ol>

<h2 id="Implementation_differences"><span class="alert">Implementation differences - TODO</span></h2>
<ul>
	<li><span class="alert">Implementation does not honor the restriction on goto statements and targets (no intervening declarations).</span></li>
	<li><span class="alert">Method expressions are partially implemented.</span></li>
	<li><span class="alert">Gccgo: allows only one init() function per source file.</span></li>
	<li><span class="alert">Gccgo: Deferred functions cannot access the surrounding function's result parameters.</span></li>
	<li><span class="alert">Gccgo: Function results are not addressable.</span></li>
	<li><span class="alert">Gccgo: Recover is not implemented.</span></li>
	<li><span class="alert">Gccgo: The implemented version of panic differs from its specification.</span></li>
</ul>