Hi guys!

Help a newbie please.

Playing with D and trying to understand some features.
Here is my try to carry out my code from C++ project to D

struct Sigmoid(T)
{
  const T Function(T value)
  {
   ...
  }
  const T DerivateFunction(const T value)
  {
   ...
  }
}

struct Neurons_layer(T = float, size_t neurons_num = 0, F = Sigmoid!T)
  if(isFloatingPoint!T && is(typeof(F.Function)))
    {
    private:
      static if(neurons_num > 0)
	T[neurons_num] _neurons_arr;
      else
	T[] _neurons_arr;

    private:
      alias Function = F.Function;
}

unittest
{
  Neurons_layer!(float,5,Sigmoid!float) nf;
}


The question is - How to make in pretty way this line:
 Neurons_layer!(float,5,Sigmoid!float) nf;

On Thursday, 14 July 2016 at 19:27:14 UTC, Andrey wrote:
> Hi guys!
>
> Help a newbie please.
>
> Playing with D and trying to understand some features.
> Here is my try to carry out my code from C++ project to D
>
> struct Sigmoid(T)
> {
>   const T Function(T value)
>   {
>    ...
>   }
>   const T DerivateFunction(const T value)
>   {
>    ...
>   }
> }
>
> struct Neurons_layer(T = float, size_t neurons_num = 0, F = Sigmoid!T)
>   if(isFloatingPoint!T && is(typeof(F.Function)))
>     {
>     private:
>       static if(neurons_num > 0)
> 	T[neurons_num] _neurons_arr;
>       else
> 	T[] _neurons_arr;
>
>     private:
>       alias Function = F.Function;
> }
>
> unittest
> {
>   Neurons_layer!(float,5,Sigmoid!float) nf;
> }
>
>
> The question is - How to make in pretty way this line:
>  Neurons_layer!(float,5,Sigmoid!float) nf;

to something like - Neurons_layer!(float,5,Sigmoid) nf;

On Thursday, 14 July 2016 at 19:28:23 UTC, Andrey wrote:
> On Thursday, 14 July 2016 at 19:27:14 UTC, Andrey wrote:
>> Hi guys!
>>
>> Help a newbie please.
>>
>> Playing with D and trying to understand some features.
>> Here is my try to carry out my code from C++ project to D
>>
>> struct Sigmoid(T)
>> {
>>   const T Function(T value)
>>   {
>>    ...
>>   }
>>   const T DerivateFunction(const T value)
>>   {
>>    ...
>>   }
>> }
>>
>> struct Neurons_layer(T = float, size_t neurons_num = 0, F = Sigmoid!T)
>>   if(isFloatingPoint!T && is(typeof(F.Function)))
>>     {
>>     private:
>>       static if(neurons_num > 0)
>> 	T[neurons_num] _neurons_arr;
>>       else
>> 	T[] _neurons_arr;
>>
>>     private:
>>       alias Function = F.Function;
>> }
>>
>> unittest
>> {
>>   Neurons_layer!(float,5,Sigmoid!float) nf;
>> }
>>
>>
>> The question is - How to make in pretty way this line:
>>  Neurons_layer!(float,5,Sigmoid!float) nf;
>
> to something like - Neurons_layer!(float,5,Sigmoid) nf;

You don't need Sigmoid!float at all. This will work:

Neurons_layer!(float, 5) nf;

as you provided a default value for the third argument.

On Thursday, 14 July 2016 at 19:28:23 UTC, Andrey wrote:
> On Thursday, 14 July 2016 at 19:27:14 UTC, Andrey wrote:
[...]
>> struct Sigmoid(T)
>> {
[...]
>> }
>>
>> struct Neurons_layer(T = float, size_t neurons_num = 0, F = Sigmoid!T)
>>   if(isFloatingPoint!T && is(typeof(F.Function)))
>>     {
[...]
>>     private:
>>       alias Function = F.Function;
>> }
>>
>> unittest
>> {
>>   Neurons_layer!(float,5,Sigmoid!float) nf;
>> }
>>
>>
>> The question is - How to make in pretty way this line:
>>  Neurons_layer!(float,5,Sigmoid!float) nf;
>
> to something like - Neurons_layer!(float,5,Sigmoid) nf;

Make F an alias parameter:

----
struct Neurons_layer(T = float, size_t neurons_num = 0, alias F = Sigmoid)
  if(isFloatingPoint!T && is(typeof(F!T.Function)))
{
    ...
    private:
      alias Function = F!T.Function;
}

unittest
{
  Neurons_layer!(float,5,Sigmoid) nf;
}
----

On 7/14/16 3:28 PM, Andrey wrote:
> On Thursday, 14 July 2016 at 19:27:14 UTC, Andrey wrote:

>> struct Neurons_layer(T = float, size_t neurons_num = 0, F = Sigmoid!T)
>>   if(isFloatingPoint!T && is(typeof(F.Function)))
>>     {
>>     private:
>>       static if(neurons_num > 0)
>>     T[neurons_num] _neurons_arr;
>>       else
>>     T[] _neurons_arr;
>>
>>     private:
>>       alias Function = F.Function;
>> }
>>
>> unittest
>> {
>>   Neurons_layer!(float,5,Sigmoid!float) nf;
>> }
>>
>>
>> The question is - How to make in pretty way this line:
>>  Neurons_layer!(float,5,Sigmoid!float) nf;
>
> to something like - Neurons_layer!(float,5,Sigmoid) nf;

You can't pass an uninstantiated template type as a type, but you can pass as an alias:

struct Neurons_layer(T = float, size_t neurons_num = 0, alias FTemp = Sigmoid)
{
   alias F = FTemp!T;
   ...

-Steve

On Thursday, 14 July 2016 at 19:48:20 UTC, Lodovico Giaretta wrote:

>
> You don't need Sigmoid!float at all. This will work:
>
> Neurons_layer!(float, 5) nf;
>
> as you provided a default value for the third argument.

Yes, default init is present, but double, real types are desirable

On Thursday, 14 July 2016 at 19:48:38 UTC, ag0aep6g wrote:

>
> Make F an alias parameter:
>
> ----
> struct Neurons_layer(T = float, size_t neurons_num = 0, alias F = Sigmoid)
>   if(isFloatingPoint!T && is(typeof(F!T.Function)))
> {
>     ...
>     private:
>       alias Function = F!T.Function;
> }
>
> unittest
> {
>   Neurons_layer!(float,5,Sigmoid) nf;
> }
> ----

WOW! It's works. Thanks.
I like D more and more )))