This problem might have been raised before, but I can't why this doesn't compile (except the last line), while a non-template version of the same code works well. Is this a bug?

---
module test;

template Graph(T)
{
    alias T[][T] Graph;
}

void add_edge(T)(ref Graph!(T) graph, T source, T target)
{
    graph[source] ~= target;
}

void main()
{
    Graph!(string) graph;

    graph.add_edge("A", "B");           // Error
    graph.add_edge!(string)("A", "B");  // Error
    add_edge(graph, "A", "B");          // Error
    add_edge!(string)(graph, "A", "B"); // OK
}
---

DMD 2.040, Mac OS X

Thanks,
Nicolas

biozic:

> This problem might have been raised before, but I can't why this doesn't compile (except the last line), while a non-template version of the same code works well. Is this a bug?

I don't know why your code doesn't work, someone more expert than me can answer you. Maybe it's a bug. In the meantime this works:

template Graph(T) {
    alias T[][T] Graph;
}

void add_edge(G, T)(ref G graph, T source, T target) if (is(G == Graph!T)) {
    graph[source] ~= target; // this is a slow operation
}

void main() {
    Graph!(string) graph;
    graph.add_edge("A", "B");
}

Bye,
bearophile

Le 08/03/10 01:27, bearophile a écrit :
> I don't know why your code doesn't work, someone more expert than me can answer you. Maybe it's a bug.
> In the meantime this works:

Yes, it works, thanks! But I believe this is just a workaround hack.

> graph[source] ~= target; // this is a slow operation

Yes, it's not the smartest graph implementation for D anyway :)


Nicolas

biozic:
> Yes, it works, thanks! But I believe this is just a workaround hack.

It's a workaround but it's not a hack. If no one else answers you then you can add a bug to bugzilla... (I am not sure it's a real bug).

Bye,
bearophile

bearophile Wrote:
> biozic:
> > Yes, it works, thanks! But I believe this is just a workaround hack.
> 
> It's a workaround but it's not a hack. If no one else answers you then you can add a bug to bugzilla... (I am not sure it's a real bug).

Yes, you're right.
I think the problem can eventually be reduced to the following simple things:
---
module test;

void foo(T)(T[] array) {}

template Type(T) { alias T Type; }
void bar(T)(Type!T t) {}

void main()
{
    int[] a = [1, 2, 3];
    foo(a); // OK
    foo!int(a); // OK
    a.foo(); // OK
    //a.foo!int(); // Error: foo(a) isn't a template

    int b = 1;
    bar!int(b); // OK
    bar(b); // Error (template deduction fails)
}
---

The first error looks like a bug.
The second one: I don't know if the template deduction engine is supposed to resolve this.

Nicolas:
> The first error looks like a bug.

Or maybe it's a limit of the compiler, it can even be a desired limit of the syntax. I don't like to write that.


> The second one: I don't know if the template deduction engine is supposed to resolve this.<

I think the answer is positive.
So I suggest you to file two bugs (even if the first one is not a bug, it can be useful to add a note to the D docs that explain this syntax is not allowed).

Bye,
bearophile

Nicolas:
> The first error looks like a bug.
> The second one: I don't know if the template deduction engine is supposed to resolve this.

For now I have added the second only: http://d.puremagic.com/issues/show_bug.cgi?id=3904

Bye,
bearophile

Le 09/03/10 00:28, bearophile a écrit :
> Nicolas:
>> The first error looks like a bug.
>> The second one: I don't know if the template deduction engine is supposed to resolve this.
>
> For now I have added the second only:
> http://d.puremagic.com/issues/show_bug.cgi?id=3904

Thanks. I've filed the first one, just in case.
Bye,
Nicolas