If I have a struct like this
type myStruct struct {
mystring string
myint int
}
and if I have a function that returns a new myStruct like this
func New() myStruct {
s := myStruct{}
s.mystring = "string"
s.myint = 1
return s
}
Because I first store it in the "s" variable before returning it, is my function actually making 2 myStruct values instead of one?
And if so, is it then a better practice to make sure I don't first store it in the variable?
The return statement will return a copy of the myStruct object value. If it is a small object then this is fine.
If you intend for the caller to be able to modify this object, and the struct will have methods that use a pointer as the receiver, then it makes more sense to return a pointer to your struct instead:
func New() *myStruct {
s := myStruct{}
s.mystring = "string"
s.myint = 1
return &s
}
You can see the copy happening when you compare the memory address of value vs pointer return types: http://play.golang.org/p/sj6mivYSHg
package main
import (
"fmt"
)
type myStruct struct {
mystring string
myint int
}
func NewObj() myStruct {
s := myStruct{}
fmt.Printf("%p\n", &s)
return s
}
func NewPtr() *myStruct {
s := &myStruct{}
fmt.Printf("%p\n",s)
return s
}
func main() {
o := NewObj()
fmt.Printf("%p\n",&o)
p := NewPtr()
fmt.Printf("%p\n",p)
}
0xf8400235a0 // obj inside NewObj()
0xf840023580 // obj returned to caller
0xf840023640 // ptr inside of NewPtr()
0xf840023640 // ptr returned to caller
I'm definitely not a Go expert (or even novice :) ), but as #max.haredoom mentioned, you can allocate variables in the function signature itself. In that way, you can also omit the s in the return:
package main
import "fmt"
type myStruct struct {
mystring string
myint int
}
func New() (s myStruct) {
s.mystring = "string"
s.myint = 1
return
}
func main() {
r := New()
fmt.Println(r)
}
// Outputs {string 1}
In the examples that I have come across in Effective Go, it does seem to be the most common way of doing things of this nature, but again, I am definitely not an authority on the subject (and will look for additional info on the actual performance).
I think I found the answer by using defer.
I updated the function so that there's a deferred modification to the myStruct value. This means it will happen after the return, but before it is received on the other end.
When I do this, the struct that is received by the caller does not show the updated value, so it appears as though I am indeed returning a copy.
func New() myStruct {
s := myStruct{}
defer func() {
s.mystring = "new value" // defer an update to the value
}()
s.mystring = "string"
s.myint = 1
return s
}
func main() {
b := New()
fmt.Println(b) // still shows the original value
}
http://play.golang.org/p/WWQi8HpDny
Related
Below I have an example of one structure which embeds another. I'm trying to figure out how to pass the more specific structure pointer to be stored in a less specific one. You can think of it as a collection. Wrapping in an interface doesn't seem to work, as doing so would make a copy, which isn't valid for structs with locks. Ideas?
package stackoverflow
import "sync"
type CoolerThingWithLock struct {
fancyStuff string
ThingWithLock
}
func NewCoolerThingWithLock() *CoolerThingWithLock {
coolerThingWithLock := &CoolerThingWithLock{}
coolerThingWithLock.InitThingWithLock()
return coolerThingWithLock
}
type ThingWithLock struct {
value int
lock sync.Mutex
children []*ThingWithLock
}
func (thingWithLock *ThingWithLock) InitThingWithLock() {
thingWithLock.children = make([]*ThingWithLock, 0)
}
func NewThingWithLock() *ThingWithLock {
newThingWithLock := &ThingWithLock{}
newThingWithLock.InitThingWithLock()
return newThingWithLock
}
func (thingWithLock *ThingWithLock) AddChild(newChild *ThingWithLock) {
thingWithLock.children = append(thingWithLock.children, newChild)
}
func (thingWithLock *ThingWithLock) SetValue(newValue int) {
thingWithLock.lock.Lock()
defer thingWithLock.lock.Unlock()
thingWithLock.value = newValue
for _, child := range thingWithLock.children {
child.SetValue(newValue)
}
}
func main() {
thingOne := NewThingWithLock()
thingTwo := NewCoolerThingWithLock()
thingOne.AddChild(thingTwo)
thingOne.SetValue(42)
}
Error: cannot use thingTwo (type *CoolerThingWithLock) as type
*ThingWithLock in argument to thingOne.AddChild
It's impossible to store the wrapping type in []*ThignWithLock since go has no notion of structural subtyping.
Your assertion that an interface will result in copying is incorrect, and you can get the desired effect by doing:
type InterfaceOfThingThatParticipatesInAHierarchy interface {
AddChild(InterfaceOfThingThatParticipatesInAHierarchy)
SetValue(int)
}
type ThingWithLock struct {
...
children []InterfaceOfThingThatParticipatesInAHierarchy
}
func (thingWithLock *ThingWithLock) AddChild(newChild InterfaceOfThingThatParticipatesInAHierarchy) { ... }
As long as the interface is implemented on a *ThingWithLock and not ThingWithLock, there will be no copying of the receiver struct itself, only the pointer to the struct will be copied on the stack.
I'm trying to dynamically call functions returning different types of struct.
For example, let's take the following code.
struct A {
Name string
Value int
}
struct B {
Name1 string
Name2 string
Value float
}
func doA() (A) {
// some code returning A
}
func doB() (B) {
// some code returning B
}
I would like to pass either the function doA or doB as an argument to a generic function that would execute the function and JSON-encode the result. Like the following:
func Generic(w io.Writer, fn func() (interface {}) {
result := fn()
json.NewEncoder(w).Encode(result)
}
But when I do:
Generic(w, doA)
I get the following error:
cannot use doA (type func() (A)) as type func() (interface {})
Is there a way to achieve this dynamic call?
First, let me remark that func() (interface{}) means the same thing as func() interface{}, so I'll use the shorter form.
Passing a function of type func() interface{}
You can write a generic function that takes a func() interface{} argument as long as the function that you pass to it has type func() interface{}, like this:
type A struct {
Name string
Value int
}
type B struct {
Name1 string
Name2 string
Value float64
}
func doA() interface{} {
return &A{"Cats", 10}
}
func doB() interface{} {
return &B{"Cats", "Dogs", 10.0}
}
func Generic(w io.Writer, fn func() interface{}) {
result := fn()
json.NewEncoder(w).Encode(result)
}
You can try out this code in a live playground:
http://play.golang.org/p/JJeww9zNhE
Passing a function as an argument of type interface{}
If you want to write functions doA and doB that return concretely typed values, you can pass the chosen function as an argument of type interface{}. Then you can use the reflect package to make a func() interface{} at run-time:
func Generic(w io.Writer, f interface{}) {
fnValue := reflect.ValueOf(f) // Make a concrete value.
arguments := []reflect.Value{} // Make an empty argument list.
fnResults := fnValue.Call(arguments) // Assume we have a function. Call it.
result := fnResults[0].Interface() // Get the first result as interface{}.
json.NewEncoder(w).Encode(result) // JSON-encode the result.
}
More concisely:
func Generic(w io.Writer, fn interface{}) {
result := reflect.ValueOf(fn).Call([]reflect.Value{})[0].Interface()
json.NewEncoder(w).Encode(result)
}
Complete program:
package main
import (
"encoding/json"
"io"
"os"
"reflect"
)
type A struct {
Name string
Value int
}
type B struct {
Name1 string
Name2 string
Value float64
}
func doA() *A {
return &A{"Cats", 10}
}
func doB() *B {
return &B{"Cats", "Dogs", 10.0}
}
func Generic(w io.Writer, fn interface{}) {
result := reflect.ValueOf(fn).Call([]reflect.Value{})[0].Interface()
json.NewEncoder(w).Encode(result)
}
func main() {
Generic(os.Stdout, doA)
Generic(os.Stdout, doB)
}
Live playground:
http://play.golang.org/p/9M5Gr2HDRN
Your return signature is different for these functions:
fn func() (interface {}) vs. func doA() (A) and func doB() (B)
You are getting a compiler error because you are passing a function with a different signature into your Generic function. To address this issue you can change your functions to return interface{}.
This is an example of how to do that, I am using anonymous structs and printing the return value out rather than serializing them but this applies just the same to your example:
package main
import "fmt"
func doA() interface{} {
return struct {
Name string
Value int
}{
"something",
5,
}
}
func doB() interface{} {
return struct {
Name1 string
Name2 string
Value float64
}{
"something",
"or other",
5.3,
}
}
func main() {
fmt.Println("Hello, playground", doA(), doB())
}
Experiment with this in the Go Playground: http://play.golang.org/p/orrJw2XMW8
If I have an interface being passed into a function, is there a way to tell if the item passed in is a struct or a pointer to a struct? I wrote this silly test to illustrate what I need to figure out.
type MyStruct struct {
Value string
}
func TestInterfaceIsOrIsntPointer(t *testing.T) {
var myStruct interface{} = MyStruct{Value: "hello1"}
var myPointerToStruct interface{} = &MyStruct{Value: "hello1"}
// the IsPointer method doesn't exist on interface, but I need something like this
if myStruct.IsPointer() || !myPointerToStruct.IsPointer() {
t.Fatal("expected myStruct to not be pointer and myPointerToStruct to be a pointer")
}
}
func isStruct(i interface{}) bool {
return reflect.ValueOf(i).Type().Kind() == reflect.Struct
}
You can test via changing type according to your needs such as reflect.Ptr. You can even get pointed value with reflect.Indirect(reflect.ValueOf(i)) after you ensured it's a pointer.
Addition:
It seems reflect.Value has a Kind method so reflect.ValueOf(i).Kind() is enough.
You could use the reflect package:
i := 42
j := &i
kindOfJ := reflect.ValueOf(j).Kind()
fmt.Print(kindOfJ == reflect.Ptr)
If you know the "real" type of the interface, you can simply use a type switch:
type MyStruct struct {
Value string
}
func TestInterfaceIsOrIsntPointer(t *testing.T) {
var myStruct interface{} = MyStruct{Value: "hello1"}
var myPointerToStruct interface{} = &MyStruct{Value: "hello1"}
// the IsPointer method doesn't exist on interface, but I need something like this
switch myStruct.(type) {
case MyStruct:
// ok
break
case *MyStruct:
// error here
break
}
switch myPointerToStruct.(type) {
case MyStruct:
// error here
break
case *MyStruct:
// ok
break
}
}
The code is longer but at least you don't need to use the reflect package.
I try to convert interface{} to struct person...
package main
import (
"encoding/json"
"fmt"
)
func FromJson(jsonSrc string) interface{} {
var obj interface{}
json.Unmarshal([]byte(jsonSrc), &obj)
return obj
}
func main() {
type person struct {
Name string
Age int
}
json := `{"Name": "James", "Age": 22}`
actualInterface := FromJson(json)
fmt.Println("actualInterface")
fmt.Println(actualInterface)
var actual person
actual = actualInterface // error fires here -------------------------------
// -------------- type assertion always gives me 'not ok'
// actual, ok := actualInterface.(person)
// if ok {
// fmt.Println("actual")
// fmt.Println(actual)
// } else {
// fmt.Println("not ok")
// fmt.Println(actual)
// }
}
... But got error:
cannot use type interface {} as type person in assignment: need type assertion
To solve this error I tried to use type assertion actual, ok := actualInterface.(person) but always got not ok.
Playground link
The usual way to handle this is to pass a pointer to the output value to your decoding helper function. This avoids type assertions in your application code.
package main
import (
"encoding/json"
"fmt"
)
func FromJson(jsonSrc string, v interface{}) error {
return json.Unmarshal([]byte(jsonSrc), v)
}
func main() {
type person struct {
Name string
Age int
}
json := `{"Name": "James", "Age": 22}`
var p person
err := FromJson(json, &p)
fmt.Println(err)
fmt.Println(p)
}
Your problem is that you're creating an empty interface to begin with, and telling json.Unmarshal to unmarshal into it. While you've defined a person type, json.Unmarshal has no way of knowing that that's what you intend the type of the JSON to be. To fix this, move the definition of person to the top level (that is, move it out of the body of main), and changeFromJson` to this:
func FromJson(jsonSrc string) interface{} {
var obj person{}
json.Unmarshal([]byte(jsonSrc), &obj)
return obj
}
Now, when you return obj, the interface{} that's returned has person as its underlying type. You can run this code on the Go Playground.
By the way, your code is a bit un-idiomatic. I left the original Playground link unmodified except for my corrections so that it wouldn't be needlessly confusing. If you're curious, here's a version that's cleaned up to be more idiomatic (including comments on why I made the changes I did).
When I define function
func test(a int, b int) int {
//bla
}
I must set arguments and return value types. How I can return value based on argument type, ex
func test(argument type) type {
//if argument type == string, must return string
//or else if argument int, must return integer
}
Can I do this and how?
Go lacks generics, (not going to argue this point one way or the other), you can achieve this by passing interface{} to functions and then doing a type assertion on the other side.
package main
import "fmt"
func test(t interface{}) interface{} {
switch t.(type) {
case string:
return "test"
case int:
return 54
}
return ""
}
func main() {
fmt.Printf("%#v\n", test(55))
fmt.Printf("%#v", test("test"))
}
You will have to type assert the value you get out
v := test(55).(int)
Go does not yet have generics like C# or Java.
It does have an empty interface (interface{})
Here is code that I believe answers your question, if I understood it correctly:
package main
import (
"fmt"
"reflect"
)
type generic interface{} // you don't have to call the type generic, you can call it X
func main() {
n := test(10) // I happen to pass an int
fmt.Println(n)
}
func test(arg generic) generic {
// do something with arg
result := arg.(int) * 2
// check that the result is the same data type as arg
if reflect.TypeOf(arg) != reflect.TypeOf(result) {
panic("type mismatch")
}
return result;
}