D template basics: argument deduction and value types

Introduction

So you've started writing D code and are happy writing simple templates like the one below to calculate dot product.

//value.d
import std.stdio: writeln;
import std.traits: isFloatingPoint;

/*
  To compile:
  dmd value.d && ./value
*/

T dot(T)(T[] x, T[] y)
if(isFloatingPoint!T)
{
  T ret = 0;
  foreach(i; 0..x.length)
    ret += x[i]*y[i];
  return ret;
}

void main()
{
  auto x = [1.0, 2.0, 3.0];
  writeln("dot: ", x.dot(x));
}

Template argument deduction

Now you need write something a bit more complicated for example to get the number of elements in a static array. You can do this by decomposing the template types using templates or template functions to obtain the number of elements in your static array. The example below does this using a template function.

//decomp.d
import std.stdio: writeln;

/*
  //To compile and run:
  rdmd decomp.d
*/

auto getStaticArraySize(A: T[N], T, alias N)(A arr)
{
  return N;
}

void main()
{
  long[4] arr = [1L, 2L, 3L, 4L];
  writeln("Static Array Size: ", getStaticArraySize(arr));
}

Here the keyword alias has to be used when you are not referring to a type, since N is a number rather than a type we have to use the alias keyword to differentiate it as such. You can also do this with a template:

//decompT.d
import std.stdio: writeln;

/*
  //To compile and run:
  rdmd decompT.d
*/

template getStaticArraySize(A: T[N], T, alias N)
{
  enum getStaticArraySize = N;
}

void main()
{
  long[4] arr = [1L, 2L, 3L, 4L];
  writeln("Static Array Size using template: ", getStaticArraySize!(typeof(arr)));
}

Value types

The terminology here is a bit confusing, a common meaning of "value type" differentiates value and reference variables. That's not what this is about, here we are talking about when a value is actually part of a type signature rather than being contained in an object ... ooo ... interesting! Consider the following:

struct Value(alias val){}

That is what I'm referring to as a "value type". Here the value val is part of the type signature of the Value template. So if we create one of these, the value in val will be available to us at compile time, but how do we get hold of val? We use an argument deduction pattern as before:

import std.stdio: writeln;
import std.traits: isFloatingPoint;

/*
  //To compile and run:
  rdmd value.d
*/

struct Value(alias val){}

auto getValue(A: V!N, alias V, alias N)(A x)
{
  return N;
}


void main()
{
  alias myValueType = Value!(4.2); // This is the type
  auto val = myValueType();
  writeln("The type signature: ", myValueType.stringof);
  writeln("Instance of the value type: ", val);
  writeln("Get the value from the type signature: ", getValue(val));
}

That's it.