www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.learn - Combining variadic functions with class templates

reply Sebastian Schuberth <sschuberth gmail.com> writes:
Hi,

I'm all new to D (coming from C++), and currently playing around with 
the language. I'm using DMD 2.049 and was expecting this to compile:

struct Vector(alias N,T)
{
     Vector(T[N] v ...) {
         data=v;
     }

     T data[N];
};

alias Vector!(3,float) Vec3f;

void main()
{
     Vec3f v(1,1,1);
}

But instead I get these errors:

Building Release\ConsoleApp1.exe...
main.d(3): found 'v' when expecting ')'
main.d(3): semicolon expected, not '...'
main.d(3): Declaration expected, not '...'
main.d(8): unrecognized declaration
Building Release\ConsoleApp1.exe failed!

What's wrong with the above code? Can member functions / constructors 
not be variadic? Is is not allowed to use template argument N in static 
array variadic function arguments?

Thanks in advance.

-- 
Sebastian Schuberth
Sep 30 2010
next sibling parent reply "Simen kjaeraas" <simen.kjaras gmail.com> writes:
Sebastian Schuberth <sschuberth gmail.com> wrote:


 Hi,

 I'm all new to D (coming from C++), and currently playing around with  
 the language. I'm using DMD 2.049 and was expecting this to compile:

 struct Vector(alias N,T)
 {
      Vector(T[N] v ...) {


This is not a constructor, but a malformed function. In D, constructors
bear the name 'this'[1].



 What's wrong with the above code? Can member functions / constructors  
 not be variadic? Is is not allowed to use template argument N in static  
 array variadic function arguments?


The error is, as said above, the name of your constructor. This leads to
a less-than-readable error message, which ultimately is trying to tell
you what I'm saying.

[1]: http://digitalmars.com/d/2.0/struct.html#Struct-Constructor

-- 
Simen
Sep 30 2010
parent reply Sebastian Schuberth <sschuberth gmail.com> writes:
On 30.09.2010 19:58, Simen kjaeraas wrote:


 This is not a constructor, but a malformed function. In D, constructors
 bear the name 'this'[1].


Doh, thanks, I should have already known that ... changing "Vector" to 
"this" makes it better, but with

struct Vector(alias N,T)
{
     this(T[N] v ...) {
         data=v;
     }

     T data[N];
};

alias Vector!(3,float) Vec3f;

void main()
{
     Vec3f v(1,1,1);
}

I still get

Building Release\ConsoleApp1.exe...
main.d(14): found 'v' when expecting ';' following statement
Building Release\ConsoleApp1.exe failed!

-- 
Sebastian Schuberth
Sep 30 2010
parent reply wrzosk <dprogr gmail.com> writes:
W dniu 2010-10-01 03:13, Sebastian Schuberth pisze:

 On 30.09.2010 19:58, Simen kjaeraas wrote:


 This is not a constructor, but a malformed function. In D, constructors
 bear the name 'this'[1].


 Doh, thanks, I should have already known that ... changing "Vector" to
 "this" makes it better, but with

 struct Vector(alias N,T)
 {
 this(T[N] v ...) {
 data=v;
 }

 T data[N];
 };

 alias Vector!(3,float) Vec3f;

 void main()
 {
 Vec3f v(1,1,1);
 }

 I still get

 Building Release\ConsoleApp1.exe...
 main.d(14): found 'v' when expecting ';' following statement
 Building Release\ConsoleApp1.exe failed!


Change
   Vec3f v(1,1,1);
into
   Vec3f v = Vec3f(1,1,1);
Sep 30 2010
parent reply Sebastian Schuberth <sschuberth gmail.com> writes:
On 30.09.2010 20:20, wrzosk wrote:


 void main()
 {
 Vec3f v(1,1,1);
 }

 I still get

 Building Release\ConsoleApp1.exe...
 main.d(14): found 'v' when expecting ';' following statement
 Building Release\ConsoleApp1.exe failed!


 Change
 Vec3f v(1,1,1);
 into
 Vec3f v = Vec3f(1,1,1);


Thanks, that works, but that syntax seems a little verbose to me. Is the 
first syntax generally not supported in D?

-- 
Sebastian Schuberth
Sep 30 2010
parent reply Pelle <pelle.mansson gmail.com> writes:
On 09/30/2010 08:25 PM, Sebastian Schuberth wrote:

 On 30.09.2010 20:20, wrzosk wrote:


 void main()
 {
 Vec3f v(1,1,1);
 }

 I still get

 Building Release\ConsoleApp1.exe...
 main.d(14): found 'v' when expecting ';' following statement
 Building Release\ConsoleApp1.exe failed!


 Change
 Vec3f v(1,1,1);
 into
 Vec3f v = Vec3f(1,1,1);


 Thanks, that works, but that syntax seems a little verbose to me. Is the
 first syntax generally not supported in D?


It is not. You can use auto v = Vec3f(1,1,1);
Sep 30 2010
parent Philippe Sigaud <philippe.sigaud gmail.com> writes:
 Thanks, that works, but that syntax seems a little verbose to me. Is the
 first syntax generally not supported in D?


 It is not. You can use auto v = Vec3f(1,1,1);


If you use a creation function, types and length are deduced
automatically, there is no need to indicate them to the compiler.
Let's add some sugar on top, it complexifies the signature a bit but
makes creating a vector a pleasure:


import std.traits;

auto vector(Ts...)(Ts values)  if (!is(CommonType!(Ts) == void))
{
    return Vector!( CommonType!(Ts), Ts.length)(values);
}

(CommonType: http://digitalmars.com/d/2.0/phobos/std_traits.html#CommonType)

usage:

auto v1 = vector(1,2,3);        // v1 is a Vector!(int,3)

auto v2 = vector(1.5);           // v2 is a Vector!(double, 1)

auto v3 = vector(1,2, 3.1415, 2.71828); // v3 is a Vector!(double, 4):
CommonType!(int,int,double,double) is double

auto v4 = vector('a','b');        // Vector!(char,2)

auto v5 = vector("abc", 3);    // Compile-time error, impossible vector.

Philippe
Sep 30 2010
prev sibling parent bearophile <bearophileHUGS lycos.com> writes:
Sebastian Schuberth:


 I'm all new to D (coming from C++), and currently playing around with 
 the language. I'm using DMD 2.049 and was expecting this to compile:
 
 struct Vector(alias N,T)
 {
      Vector(T[N] v ...) {
          data=v;
      }
 
      T data[N];
 };
 
 alias Vector!(3,float) Vec3f;
 
 void main()
 {
      Vec3f v(1,1,1);
 }


This is interesting code, thank you. Others have already shown you the two bugs
in your code, plus other ways to design it. At the end of struct/enum/class
definitions don't put a semicolon, it's not needed in D.

I have filed an enhancement request about those error messages:
http://d.puremagic.com/issues/show_bug.cgi?id=4962

Bye,
bearophile
Sep 30 2010