Hi all,

Here is a snipet:

import std.stdio;

struct A(T, T x0)
{
T x = x0;
}

int main(char[][] args)
{
alias A!(int, 8) Ai; // test.d(11)
Ai ai;
writefln(ai.x);
return 0;
}

DMD output is:
test.d(11): template instance A!(int,8) does not match any template declaration

However snipet:

import std.stdio;

struct A(T, int x0) // note int instead of T
{
T x = x0;
}

int main(char[][] args)
{
alias A!(int, 8) Ai;
Ai ai;
writefln(ai.x);
return 0;
}

works as expected. Is former is a templates issue or bug?

--
Victor Nakoryakov (aka nail)
nail-mail[at]mail.ru

Krasnoznamensk, Moscow, Russia

Victor Nakoryakov schrieb am 2006-03-11:
> Hi all,
>
> Here is a snipet:
>
> import std.stdio;
>
> struct A(T, T x0)
> {
> T x = x0;
> }
>
> int main(char[][] args)
> {
> alias A!(int, 8) Ai; // test.d(11)
> Ai ai;
> writefln(ai.x);
> return 0;
> }
>
> DMD output is:
> test.d(11): template instance A!(int,8) does not match any template declaration
>
> However snipet:
>
> import std.stdio;
>
> struct A(T, int x0) // note int instead of T
> {
> T x = x0;
> }
>
> int main(char[][] args)
> {
> alias A!(int, 8) Ai;
> Ai ai;
> writefln(ai.x);
> return 0;
> }
>
> works as expected. Is former is a templates issue or bug?

Added to DStress as http://dstress.kuehne.cn/run/t/template_31_A.d http://dstress.kuehne.cn/run/t/template_31_B.d http://dstress.kuehne.cn/run/t/template_31_C.d http://dstress.kuehne.cn/run/t/template_31_D.d

template_31_C seems funny, but I'm not aware of any statement against it.

Thomas

Thomas Kuehne wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Victor Nakoryakov schrieb am 2006-03-11:
>> Hi all,
>>
>> Here is a snipet:
>>
>> import std.stdio;
>>
>> struct A(T, T x0)
>> {
>> T x = x0;
>> }
>>
>> int main(char[][] args)
>> {
>> alias A!(int, 8) Ai; // test.d(11)
>> Ai ai;
>> writefln(ai.x);
>> return 0;
>> }

If that had worked, it would have been a new feature. You're obtaining a type AND a value! Note that value parameters only work for basic types and char[]. It would be very cool, though -- it would enable a lot of code reuse for metaprogramming.

For example, pow!(real z, int n) and pow!(creal z, int n) would be able to share a common implementation. The fact that they can't has been bugging me for some time.

Don Clugston wrote:
> Thomas Kuehne wrote:
>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>>
>> Victor Nakoryakov schrieb am 2006-03-11:
>>> Hi all,
>>>
>>> Here is a snipet:
>>>
>>> import std.stdio;
>>>
>>> struct A(T, T x0)
>>> {
>>> T x = x0;
>>> }
>>>
>>> int main(char[][] args)
>>> {
>>> alias A!(int, 8) Ai; // test.d(11)
>>> Ai ai;
>>> writefln(ai.x);
>>> return 0;
>>> }
> 
> If that had worked, it would have been a new feature. You're obtaining a type AND a value! Note that value parameters only work for basic types and char[]. It would be very cool, though -- it would enable a lot of code reuse for metaprogramming.
> 
> For example, pow!(real z, int n) and pow!(creal z, int n) would be able to share a common implementation. The fact that they can't has been bugging me for some time.

For what it's worth, this is legal in C++:

    #include <iostream>

    template<class T, T x>
    struct S
    {
        static const T val = x;
    };

    int main()
    {
        int x = S<int,5>::val;
        std::cout << x << '\n';
    }

Sean