What is the right way to handle NaN in Java?

Question

For example, if you are computing the precision

p = correct / total

Would you make sure you don't divide by zero:

double p;
if (total == 0.0) {
  p = 0.0;
}
else {
  p == correct / total;
}

Or check if you get a NaN?

double p = correct / total;
if (Double.isNaN(p)) {
  p = 0.0;
}

Is there a benefit to an approach, or is it personal preference?

Answer 1

I would use the first approach, but instead of comparing to 0, I would compare the Math.abs(total) < TOLERANCE where TOLERANCE is some small value like 0.0001. This will prevent things very close to 0 from skewing results.