SliceFL Example

Here we present a short example that emphasizes the usefulness of our approach. In this example, the first code snippet contains a Circle class with some basic operations. The error is in the constructor when calculating the perimeter (a multiplier of two is missing).

public class Circle {
    private double area;
    private double perimeter;
    public Circle(double radius) {
        area = radius * radius * Math.PI;
        perimeter = radius * Math.PI; // faulty statement
    }
    double getArea() {
        return area;
    }
    double getPerimeter() {
        return perimeter;
    }
}

The second code snippet shows how to test the Circle class. If these tests are run, t2 will report an error, however, tests t1 and t3 will pass. This is a problem because t3 covers the faulty statement in the constructor and nevertheless does not fail, i.e. it will distort the suspiciousness value associated with the given row (i.e. the perimeter calculation) (you can see that it, along with several others, has a value of 0.5).

import junit.framework.TestCase;

public class CircleTest extends TestCase {
	 static Circle circle = new Circle(0);
    public void t1() {
        assertEquals(Math.PI, new Circle(1).getArea(), 1e-10);
    }
    public void t2() {
        assertEquals(2.0*Math.PI, new Circle(1).getPerimeter(), 1e-10);
    }
    public void t3() {
        assertEquals(0, circle.getPerimeter(), 1e-10);
    }
}

In contrast, with the slicing approach (as shown in the second table), it is clearly the faulty row that already gets the highest value. The reason for this is that the Circle object used here has already been created, so when the slice is calculated, the instructions in it will not be rerun and therefore will not be included in the slice (unlike the second test, where a new object is used).

Coverage-based spectrum and Fault Localization results

	4	5	6	8	9	11	12	R
C(t1)	1	1	1	1	1	0	0	0
C(t2)	1	1	1	0	0	1	1	1
C(t3)	0	0	0	0	0	1	1	0
ef	1	1	1	0	0	1	1
ep	1	1	1	1	1	1	1
nf	0	0	0	1	1	0	0
np	1	1	1	1	1	1	1
Bar	0.5	0.5	0.5	0.0	0.0	0.5	0.5

Slice-based spectrum and Fault Localization results

	4	5	6	8	9	11	12	R
DS(t1)	1	1	0	1	1	0	0	0
DS(t2)	1	0	1	0	0	1	1	1
DS(t3)	0	0	0	0	0	1	1	0
ef	1	0	1	0	0	1	1
ep	1	1	0	1	1	1	1
nf	0	1	0	1	1	0	0
np	1	1	2	1	1	1	1
Bar	0.5	0.0	1.0	0.0	0.0	0.5	0.5