Is an empty class in some cases correct

Question

I know it's generally a design flaw to have an empty class, yet I'm wondering whether there are some cases where it isn't to have an empty (abstract) class to group objects. Especially when polymorphism is needed.

For example I have a group location. They can consist of an area, a line (a road for example) or a specific point (such as a house). In the data structure it only requires to contain the coordinates.

I know they all have at least a single point in common, yet it's really weird property wise when I would have a coordinate at my location class.

At the moment I was thinking of the following solution:

public abstract class Location {

}

public class Point: Location 
{
    public double X {get;set;}
    public double Y {get;set;}
}

public class Line : Location
{
    public double StartX {get;set;}
    public double StartY {get;set;}
    public double EndX {get;set;}
    public double EndY {get;set;}
}

I know it's not a perfect example as there is another more correct solution to this case and there are probably objects for this solution found within the .Net libraries.

But how are things like this solved overall?

Answer 1

Polymorphism with no shared interface will ONLY let you store objects in the same (strongly typed) container, not actually DO anything. You can't disptach method calls or anything like that.

In your specific example, I think composition would be more appropriate, no? You already have an abstraction for a specific point in 2d space, so why woudn't you reuse it in your abstraction of all points between two points?

public class Point
{
    public double X { get; }
    public double Y { get; }

    public Point(int x, int y)
    {
        X = x;
        Y = y;
    }
}

public class Line
{
    public Point Start { get; }
    public Point End { get; }

    public Line(Point start, Point end)
    {
        Start = start;
        End = end;
    }
}

EDIT: I'd also mention, having a Line inherit from Location is just misleading. A line in the mathematical (euclidean) sense is a set of points defined by two points. Of course, if what you're interested in are JUST the points BETWEEN the start/end points, then what you have isn't a line, but a line segment.

Answer 2

The two subclasses does not have any interface in common, so why you do you want to have them inherit from a common class in the first place? The Location class does not provide you any benefit as far as I can tell.

To me it would seem to more logical to have the Line contain two Point's.