Although Go absorbs many features from all kinds of other languages, Go is mainly viewed as a C family language. One evidence is Go also supports pointers. Go pointers and C pointers are much similar in many aspects, but there are also some differences between Go pointers and C pointers. This article will list all kinds of concepts and details related to pointers in Go.
A memory address means a specific memory location in programming.
Generally, a memory address is stored as an unsigned native (integer) word. The size of a native word is 4 (bytes) on 32-bit architectures and 8 (bytes) on 64-bit architectures. So the theoretical maximum memory space size is 232 bytes, a.k.a. 4GB (1GB == 230 bytes), on 32-bit architectures, and is 264 bytes a.k.a 16EB (1EB == 1024PB, 1PB == 1024TB, 1TB == 1024GB), on 64-bit architectures.
Memory addresses are often represented with hex integer literals,
such as 0x1234CDEF
.
The address of a value means the start address of the memory segment occupied by the direct part of the value.
Pointer is one kind of type in Go. A pointer value is used to store a memory address, which is generally the address of another value.
Unlike C language, for safety reason, there are some restrictions made for Go pointers. Please read the following sections for details.
In Go, an unnamed pointer type can be represented with *T
,
where T
can be an arbitrary type.
Type T
is called the base type of pointer type *T
.
We can declare named pointer types, but generally, it’s not recommended to use named pointer types, for unnamed pointer types have better readabilities.
If the underlying type
of a named pointer type is *T
,
then the base type of the named pointer type is T
.
Two unnamed pointer types with the same base type are the same type.
Example:*int // An unnamed pointer type whose base type is int.
**int // An unnamed pointer type whose base type is *int.
// Ptr is a named pointer type whose base type is int.
type Ptr *int
// PP is a named pointer type whose base type is Ptr.
type PP *Ptr
Zero values of any pointer types are represented with the predeclared nil
.
No addresses are stored in nil pointer values.
A value of a pointer type whose base type is T
can only store the addresses of values of type T
.
In Go 101, the word "reference" indicates a relation. For example, if a pointer value stores the address of another value, then we can say the pointer value (directly) references the other value, and the other value has at least one reference. The uses of the word "reference" in Go 101 are consistent with Go specification.
When a pointer value references another value, we also often say the pointer value points to the other value.
new
function can be used to allocate memory for a value
of any type. new(T)
will allocate memory for a T
value
(an anonymous variable) and return the address of the T
value.
The allocated value is a zero value of type T
.
The returned address is viewed as a pointer value of type *T
.
t
of type T
,
we can use the expression &t
to take the address of t
,
where &
is the operator to take value addresses.
The type of &t
is viewed as *T
.
Generally speaking, an addressable value means a value which is hosted at somewhere in memory. Currently, we just need to know that all variables are addressable, whereas constants, function calls and explicit conversion results are all unaddressable. When a variable is declared, Go runtime will allocate a piece of memory for the variable. The starting address of that piece of memory is the address of the variable.
We will learn other addressable and unaddressable values from other articles later. If you have already been familiar with Go, you can read this summary to get the lists of addressable and unaddressable values in Go.
The next section will show an example on how to get pointer values.
Given a pointer value p
of a pointer type whose base type is T
,
how can you get the value at the address stored in the pointer (a.k.a., the value
being referenced by the pointer)? Just use the expression *p
,
where *
is called dereference operator.
*p
is called the dereference of pointer p
.
Pointer dereference is the inverse process of address taking.
The result of *p
is a value of type T
(the base type of the type of p
).
Dereferencing a nil pointer causes a runtime panic.
package main
import "fmt"
func main() {
p0 := new(int) // p0 points to a zero int value.
fmt.Println(p0) // (a hex address string)
fmt.Println(*p0) // 0
// x is a copy of the value at
// the address stored in p0.
x := *p0
// Both take the address of x.
// x, *p1 and *p2 represent the same value.
p1, p2 := &x, &x
fmt.Println(p1 == p2) // true
fmt.Println(p0 == p1) // false
p3 := &*p0 // <=> p3 := &(*p0) <=> p3 := p0
// Now, p3 and p0 store the same address.
fmt.Println(p0 == p3) // true
*p0, *p1 = 123, 789
fmt.Println(*p2, x, *p3) // 789 789 123
fmt.Printf("%T, %T \n", *p0, x) // int, int
fmt.Printf("%T, %T \n", p0, p1) // *int, *int
}
The following picture depicts the relations of the values used in the above program.
package main
import "fmt"
func double(x int) {
x += x
}
func main() {
var a = 3
double(a)
fmt.Println(a) // 3
}
The double
function in the above example is expected to modify
the input argument by doubling it. However, it fails. Why?
Because all value assignments, including function argument passing, are value copying in Go.
What the double
function modified is a copy (x
) of variable a
but not variable a
.
double
function is let it return
the modification result. This solution doesn't always work for all scenarios.
The following example shows another solution, by using a pointer parameter.
package main
import "fmt"
func double(x *int) {
*x += *x
x = nil // the line is just for explanation purpose
}
func main() {
var a = 3
double(&a)
fmt.Println(a) // 6
p := &a
double(p)
fmt.Println(a, p == nil) // 12 false
}
We can find that, by changing the parameter to a pointer type,
the passed pointer argument &a
and its copy x
used in the function body both reference the same value,
so the modification on *x
is equivalent to a modification on *p
, a.k.a., variable a
.
In other words, the modification in the double
function body can be reflected out of the function now.
Surely, the modification of the copy of the passed pointer argument itself
still can't be reflected on the passed pointer argument.
After the second double
function call, the local pointer
p
doesn't get modified to nil
.
In short, pointers provide indirect ways to access some values. Many languages do not have the concept of pointers. However, pointers are just hidden under other concepts in those languages.
func newInt() *int {
a := 3
return &a
}
For safety reasons, Go makes some restrictions to pointers (comparing to pointers in C language). By applying these restrictions, Go keeps the benefits of pointers, and avoids the dangerousness of pointers at the same time.
In Go, pointers can't do arithmetic operations.
For a pointer p
, p++
and p-2
are both illegal.
If p
is a pointer to a numeric value, compilers will view
*p++
is a legal statement and treat it as (*p)++
.
In other words, the precedence of the pointer dereference operator *
is higher than the increment operator ++
and decrement operator --
.
package main
import "fmt"
func main() {
a := int64(5)
p := &a
// The following two lines don't compile.
/*
p++
p = (&a) + 8
*/
*p++
fmt.Println(*p, a) // 6 6
fmt.Println(p == &a) // true
*&a++
*&*&a++
**&p++
*&*p++
fmt.Println(*p, a) // 10 10
}
T1
can be directly and explicitly converted
to another pointer type T2
only if either of the following two conditions
is get satisfied.
T1
and T2
are identical (ignoring struct tags),
in particular if either T1
and T2
is a
unnamed type
and their underlying types are identical (considering struct tags),
then the conversion can be implicit.
Struct types and values will be explained in
the next article.
T1
and T2
are both unnamed pointer types and
the underlying types of their base types are identical (ignoring struct tags).
type MyInt int64
type Ta *int64
type Tb *MyInt
the following facts exist:
*int64
can be implicitly converted to type Ta
,
and vice versa, for their underlying types are both *int64
.
*MyInt
can be implicitly converted to type Tb
,
and vice versa, for their underlying types are both *MyInt
.
*MyInt
can be explicitly converted to type *int64
,
and vice versa, for they are both unnamed and the underlying types of their base types are both int64
.
Ta
can't be directly converted to type Tb
, even if explicitly.
However, by the just listed first three facts, a value pa
of
type Ta
can be indirectly converted to type Tb
by nesting three explicit conversions, Tb((*MyInt)((*int64)(pa)))
.
None values of these pointer types can be converted to type *uint64
, in any safe ways.
==
and !=
operators.
Two Go pointer values can only be compared if either of the following three conditions
are satisfied.
nil
identifier.
package main
func main() {
type MyInt int64
type Ta *int64
type Tb *MyInt
// 4 nil pointers of different types.
var pa0 Ta
var pa1 *int64
var pb0 Tb
var pb1 *MyInt
// The following 6 lines all compile okay.
// The comparison results are all true.
_ = pa0 == pa1
_ = pb0 == pb1
_ = pa0 == nil
_ = pa1 == nil
_ = pb0 == nil
_ = pb1 == nil
// None of the following 3 lines compile ok.
/*
_ = pa0 == pb0
_ = pa1 == pb1
_ = pa0 == Tb(nil)
*/
}
The conditions to assign a pointer value to another pointer value are the same as the conditions to compare a pointer value to another pointer value, which are listed above.
As the start of this article has mentioned,
the mechanisms (specifically, the unsafe.Pointer
type)
provided by the unsafe
standard package
can be used to break the restrictions made for pointers in Go.
The unsafe.Pointer
type is like the void*
in C.
In general the unsafe ways are not recommended to use.
The Go 101 프로젝트는 Github 에서 호스팅됩니다. 오타, 문법 오류, 부정확한 표현, 설명 결함, 코드 버그, 끊어진 링크와 같은 모든 종류의 실수에 대한 수정 사항을 제출하여 Go 101을 개선을 돕는 것은 언제나 환영합니다.
주기적으로 Go에 대한 깊이 있는 정보를 얻고 싶다면 Go 101의 공식 트위터 계정인 @go100and1을 팔로우하거나 Go 101 슬랙 채널에j가입해주세요.
reflect
표준 패키지sync
표준 패키지sync/atomic
표준 패키지