On Tuesday, 5 December 2017 at 08:08:55 UTC, Daniel Kozak wrote:
You can do something like this:
interface Medoid(T) {
float distance( T other );
uint id() const @property;
}
class Item : Medoid!(Item) {
float distance( Item m ) { return 0.;}
uint id() const @property { return 1; }
}
class MedoidClassification {
this(T:Medoid!T)(T[] list) {}
//Medoid[][] getClusters() {...}
}
void main() {
auto items = new Item[10];
auto mc = new MedoidClassification( items );
}
On Tue, Dec 5, 2017 at 8:47 AM, Dirk via Digitalmars-d-learn <
digitalmars-d-learn@puremagic.com> wrote:
The distance function is implementation dependend and can only
be computed between two objects of the same class (in this
example the class is Item).
My goal is to write a module for a k-medoids clustering
algorithm. The class MedoidClassification shall be able to
partition a list of objects from the same class, which
implement the Medoid interface.
My current approach is this (which does not work):
interface Medoid {
float distance( Medoid other );
uint id() const @property;
}
class Item : Medoid {
float distance( Item m ) {...}
uint id() const @property {...}
}
class MedoidClassification {
this( Medoid[] list ) {...}
Medoid[][] getClusters() {...}
}
void main() {
Item[10] items;
auto mc = MedoidClassification( items );
}
What would be a good way to implement this?
This still defeats the purpose of having multiple Medoid types,
as each Medoid is still specified with a specific type.
