On 09/19/2015 02:33 AM, OlaOst wrote:
> Here is a class with a templated opIndex method, and an attempt to
use it:
>
> class Test
> {
> int[] numbers = [1, 2, 3];
> string[] texts = ["a", "b", "c"];
>
> Type opIndex(Type)(int index)
> {
> static if (is(Type == int))
> return numbers[index];
> static if (is(Type == string))
> return texts[index];
> }
> }
>
> void main()
> {
> auto test = new Test();
>
> auto number = test[0]!int; // does not compile, syntax error
> auto number = test!int[0]; // does not compile, syntax error
>
> int number = test[0]; // does not compile, cannot deduce type
> }
>
>
> So it is possible to define a templated opIndex method in a class, but
> is it possible to use it? If not, should it be allowed to create
> templated opIndex methods?
Templated opIndex is used for multi-dimensional array indexing. The
template arguments define the range of elements:
http://ddili.org/ders/d.en/templates_more.html#ix_templates_more.opIndex%20template
And the spec:
http://dlang.org/operatoroverloading.html#array-ops
I would prefer the following:
test.numbers[0];
test.texts[0];
but in addition to other solutions, here is another experiment:
struct indexFor(T)
{
size_t idx;
}
class Test
{
int[] numbers = [1, 2, 3];
string[] texts = ["a", "b", "c"];
T opIndex(T)(indexFor!T index)
{
static if (is(T == int))
return numbers[index.idx];
static if (is(T == string))
return texts[index.idx];
assert(false);
}
}
void main()
{
auto test = new Test();
auto number = test[indexFor!int(0)];
}
Ali
