It seems pointless to be used in primitive language constructs, as you can't specify any sort of values
func main() {
y := new([]float)
fmt.Printf("Len = %d", len(*y) ) // => Len = 0
}
For stucts it makes a b开发者_Python百科it more sense, but what's the difference between saying y := new(my_stuct)
and the seemingly more concise y := &my_struct
?
And since anything you create is based on those primitives, they will be initialized to the said zero values. So what's the point? When would you ever want to use new()
?
Sorry for the very-beginner question, but the documentation isn't always that clear.
You can't use new
for slices and maps, as in your code example, but instead you must use the make
command: make([]float, 100)
Both new(MyStruct)
and &MyStruct{}
do to the same thing, because Go will allocate values on the heap if you get their address with &. Sometimes the code just expresses it intent better in one style or the other.
Go does not have built-in support for constructors, so usually you would wrap the call to new
into a function, for example NewMyStruct()
which does all the necessary initialization. It also makes it possible to initialize private fields or hide the struct behind an interface, to prevent users of the object from directly messing with its internals. Also evolving the structure of the struct is easier that way, when you don't need to change all of its users when adding/removing/renaming/reordering fields.
make
does only work for maps, slices and channels and composite literals like type{}
work only for structs, arrays, slices, and maps. For other types, you'll have to use new
to get a pointer to a newly allocated instance (if you don't want to use a longer var v T; f(&v)
).
I guess this is useful if you want to initialize a struct:
typedef foo struct {
bar *int
}
v := foo{bar: new(int)}
精彩评论