How to have a function with a nullable string parameter in Go?

I'm used to Java's String where we can pass null rather than "" for special meanings, such as use a default value.

In Go, string is a primitive type, so I cannot pass nil (null) to a parameter that requires a string.

I could write the function using pointer type, like this:

func f(s *string)

so caller can call that function either as

f(nil)

or

// not so elegant
temp := "hello";
f(&temp) 

but the following is unfortunately not allowed:

// elegant but disallowed
f(&"hello");

What is the best way to have a parameter that receives either a string o开发者_如何学JAVAr nil?

Warning: the following is pre-Go1 code. That is, it's from a pre-release version and is not valid Go code.

I thought some more about how I would implement this using a struct. Here's what I came up with:

type MyString struct {
    val string;
}

func f(s MyString) {
    if s == nil {
        s = MyString{"some default"};
    }
    //do something with s.val
}

Then you can call f like this:

f(nil);
f(MyString{"not a default"});

If you need to handle a possibble null value (becasue you are talking to a database that may provide them for example) the database/sql package has types such as sql.NullString and sql.NullInt64 that allow you to test if you have been provided with a value or not using their .Valid field.

Not realy attend answer : but warping value in a structure can provide some generic utility methode. (Haskell Maybe ?)

//#maybe.go
package maybe

import "log"

type MayHaveValue struct {
 IsValue bool;
}

func (this MayHaveValue) IsJust() bool {
 return this.IsValue
}

type AString struct {
 MayHaveValue;
 Value string;
}

func String(aString string) AString {
 return AString{MayHaveValue{true}, aString}
}

var NoString AString = AString{MayHaveValue{false}, ""}

func (this AString) String() (value string) {
 if this.IsJust() == true {
  value = this.Value;
 } else {
  log.Crash("Access to non existent maybeString value");
 }
 return;
}

func (this AString) OrDefault(defaultString string) (value string) {
 if this.IsJust() {
  value = this.Value;
 } else {
  value = defaultString;
 }
 return;
}

//#main.go
package main

import "fmt"
import "maybe"

func say(canBeString maybe.AString) {
 if canBeString.IsJust() {
  fmt.Printf("Say : %v\n", canBeString.String());
 } else {
  fmt.Print("Nothing to say !\n");
 }
}

func sayMaybeNothing (canBeString maybe.AString) {
 fmt.Printf("Say : %v\n", canBeString.OrDefault("nothing"));
}

func main() {
 aString := maybe.String("hello");
 say(aString);
 sayMaybeNothing(aString);
 noString := maybe.NoString;
 say(noString);
 sayMaybeNothing(noString);
}

You can declare an interface to restrict the type to string, and since interfaces also accept nil, you'd have both cases covered. This is how you can implement it:

type (
    // An interface which accepts a string or a nil value.
    //
    // You can pass StrVal("text") or nil.
    StrOrNil interface{ isStrOrNil() }

    StrVal string // A string value for StrOrNil interface.
)

func (StrVal) isStrOrNil() {} // implement the interface

And this is how you use it:

func Foo(name StrOrNil) {
    switch nameVal := name.(type) {
    case StrVal:
        fmt.Printf("String value! %s\n", string(nameVal))
    default:
        fmt.Println("Null value!")
    }
}

func main() {
    Foo(StrVal("hello world"))
    Foo(nil)
}

Test it in the playground.

Loose the Java-think and just pass f(""). Then test using len():

func f(str string) { if len(str) > 0 { ... } else { ... } }

Either the string is empty and has semantic meaning of you nil case, or else has some string data to process. Can't see the problem with that.