Hello,

I'm sure this can be done, I'm just not sure what are the correct terms to search for...

Given one interface, and multiple implementation classes, I want to create a list of the classes (in compile time).
A contrived example:
===
interface Animal
{
  static const string name() @property;
  void make_noise();
}

class Dog : Animal
{
   static const string name() @property { return "dog" ; }
   void make_noise() { writeln("Woof"); }
}

class Cat : Animal
{
   static const string name() @property { return "cat" ; }
   void make_noise() { writeln("Meow"); }
}
===

What I want is:

===
static const xxxxx[] available_animals = [ dog, cat ];
===

and then:

===
foreach (a; available_animals) {
   writeln("We have a ", a.name);
}
===

Given that "name" is a static member function, it should be doable.
But I'm not sure about the correct syntax of defining "available_animals".

Any suggestions are welcomed,
 -gordon

On Sunday, 23 February 2014 at 20:34:07 UTC, Gordon wrote:
> Hello,
>
> I'm sure this can be done, I'm just not sure what are the correct terms to search for...
>
> Given one interface, and multiple implementation classes, I want to create a list of the classes (in compile time).
> A contrived example:
> ===
> interface Animal
> {
>   static const string name() @property;
>   void make_noise();
> }
>
> class Dog : Animal
> {
>    static const string name() @property { return "dog" ; }
>    void make_noise() { writeln("Woof"); }
> }
>
> class Cat : Animal
> {
>    static const string name() @property { return "cat" ; }
>    void make_noise() { writeln("Meow"); }
> }
> ===
>
> What I want is:
>
> ===
> static const xxxxx[] available_animals = [ dog, cat ];
> ===
>
> and then:
>
> ===
> foreach (a; available_animals) {
>    writeln("We have a ", a.name);
> }
> ===
>
> Given that "name" is a static member function, it should be doable.
> But I'm not sure about the correct syntax of defining "available_animals".
>
> Any suggestions are welcomed,
>  -gordon

import std.typetuple: TypeTuple;
alias available_animals = TypeTuple!(Dog, Cat);

On Sunday, 23 February 2014 at 20:52:58 UTC, anonymous wrote:
> On Sunday, 23 February 2014 at 20:34:07 UTC, Gordon wrote:
>>
>> Given one interface, and multiple implementation classes, I want to create a list of the classes (in compile time).
>
> import std.typetuple: TypeTuple;
> alias available_animals = TypeTuple!(Dog, Cat);

First, Thanks!
D is amazing :)

Now, this lead me to a weird error.
This is the complete (contrived) program:
===
import std.stdio,
	std.string,
	std.typetuple;

interface Animal
{
  static const string name() @property;
  void make_noise();
}

class Dog : Animal
{
   static const string name() @property { return "dog" ; }
   void make_noise() { writeln("woof"); }
}

class Cat : Animal
{
   static const string name() @property { return "cat" ; }
   void make_noise() { writeln("meow"); }
}

alias available_animals = TypeTuple!(Dog, Cat);

void main()
{
	string input_from_user = "dog";
	Animal a;
	foreach (i; available_animals) {
		if (i.name == input_from_user)
			a = new i;
	}

        //// -- This doesn't compile ---
	//// writeln("The ", a.name, " makes: ");
        ////

	a.make_noise();
}
===

Running the program without the marked "writeln" works fine, as expected.

Compiling with the marked writeln, produces an "underfined reference" error:
====
$ rdmd animals.d
/tmp/.rdmd-34574/rdmd-animals.d-796CFD5A46BFE9DF13BF873F65EA656C/objs/animals.o: In function `_Dmain':
animals.d:(.text._Dmain+0xdf): undefined reference to `_D7animals6Animal4nameFNdZAya'
collect2: error: ld returned 1 exit status
--- errorlevel 1
====

BTW - a work-around is simple, my question is about the theory behind this.

I understand that "Animal::name" is undefined (since "Animal" is an interface).

but if during runtime "a" has an actual type, why does the compiler looks for "Animal::name" instead of Dog::Name or Cat::Name (based on the type of "a") ?

It is likely because of the "static", but if I remove the "static" than the 'foreach' loop won't compile, because "name" is not static.

So this seems like some impossible situation...

On Sunday, 23 February 2014 at 22:14:17 UTC, Gordon wrote:
> On Sunday, 23 February 2014 at 20:52:58 UTC, anonymous wrote:
>> On Sunday, 23 February 2014 at 20:34:07 UTC, Gordon wrote:
>>>
>>> Given one interface, and multiple implementation classes, I want to create a list of the classes (in compile time).
>>
>> import std.typetuple: TypeTuple;
>> alias available_animals = TypeTuple!(Dog, Cat);
>
> First, Thanks!
> D is amazing :)
>
> Now, this lead me to a weird error.
> This is the complete (contrived) program:
> ===
> import std.stdio,
> 	std.string,
> 	std.typetuple;
>
> interface Animal
> {
>   static const string name() @property;
>   void make_noise();
> }
>
> class Dog : Animal
> {
>    static const string name() @property { return "dog" ; }
>    void make_noise() { writeln("woof"); }
> }
>
> class Cat : Animal
> {
>    static const string name() @property { return "cat" ; }
>    void make_noise() { writeln("meow"); }
> }
>
> alias available_animals = TypeTuple!(Dog, Cat);
>
> void main()
> {
> 	string input_from_user = "dog";
> 	Animal a;
> 	foreach (i; available_animals) {
> 		if (i.name == input_from_user)
> 			a = new i;
> 	}
>
>         //// -- This doesn't compile ---
> 	//// writeln("The ", a.name, " makes: ");
>         ////
>
> 	a.make_noise();
> }
> ===
>
> Running the program without the marked "writeln" works fine, as expected.
>
> Compiling with the marked writeln, produces an "underfined reference" error:
> ====
> $ rdmd animals.d
> /tmp/.rdmd-34574/rdmd-animals.d-796CFD5A46BFE9DF13BF873F65EA656C/objs/animals.o: In function `_Dmain':
> animals.d:(.text._Dmain+0xdf): undefined reference to `_D7animals6Animal4nameFNdZAya'
> collect2: error: ld returned 1 exit status
> --- errorlevel 1
> ====
>
> BTW - a work-around is simple, my question is about the theory behind this.
>
> I understand that "Animal::name" is undefined (since "Animal" is an interface).
>
> but if during runtime "a" has an actual type, why does the compiler looks for "Animal::name" instead of Dog::Name or Cat::Name (based on the type of "a") ?

static methods are statically dispatched. Since they can called without an instance, they cannot be dispatched dynamically in general.

> It is likely because of the "static", but if I remove the "static" than the 'foreach' loop won't compile, because "name" is not static.
Correct.

> So this seems like some impossible situation...
A non static method that forwards to the static one or comparison of the user input to the class names come to mind. Or use some instance of i to call (the non-static) method name.

On Sun, 23 Feb 2014 17:14:16 -0500, Gordon <me@home.com> wrote:

> Compiling with the marked writeln, produces an "underfined reference" error:
> ====
> $ rdmd animals.d
> /tmp/.rdmd-34574/rdmd-animals.d-796CFD5A46BFE9DF13BF873F65EA656C/objs/animals.o: In function `_Dmain':
> animals.d:(.text._Dmain+0xdf): undefined reference to `_D7animals6Animal4nameFNdZAya'
> collect2: error: ld returned 1 exit status
> --- errorlevel 1
> ====
>
> BTW - a work-around is simple, my question is about the theory behind this.
>
> I understand that "Animal::name" is undefined (since "Animal" is an interface).

Note that Animal.name (Animal::name is C++, we use dots in D) does not define any requirements for derived mechanisms. And calling it on a base class or interface instance does not call the derived class' version.

You should not make it static, it should be virtual. Otherwise, it will not be a virtual call. The reason you get the error is because it's trying to call Animal.name, and you haven't defined it.

> It is likely because of the "static", but if I remove the "static" than the 'foreach' loop won't compile, because "name" is not static.

You don't need the whole static name business, D is better than that :)

string input_from_user = "Dog"; // must be exact for this example, but you could use case insensitive compare if you wanted

...

if(i.stringof == input_from_user)

...

-Steve