RE: assertEquals("NaN", x, y, 0.0) where x=y=Double.NaN

> -----Original Message-----
> From: Morten Grum, PH-Consult [mailto:mg@xxxxxx]
...
> Hi
> 
> x and y are both simple double.
> 
> Oughtn't assertEquals("x and y", x, y, 0.0) succeed when both 
> x and y are
> Double.NaN?
> 
> At the moment I'm having to make a situation specific assert 
> statement when
> I expect y is Double.NaN.
> 
> I'm having to write ...
> assertTrue("When y is NaN", Double.isNaN(y));
> 
> I know java returns false on (Double.NaN==y) even when y is 
> Double.NaN, but
> for our assert statements I'd say it'd be more practical/general if
> assertEquals succeeded when both were NaN.
> 
...
They are correct, in that f0(x)=(5.0-x)/(5.0-x) and
f1(x)=(5.0-x)*(5.0-x)/(5.0-x) are not equal for
x = 5.O, and both are indeterminated (NaN). If you would change
assertEquals(double, double, double)
errors could be introduced for other users.

Why not introduce a method (in an abstract super class perhaps); tailored to
your specific needs ?
   assertEqualsNaN(double x, double y, double delta){
      if(Double.isNaN(x) && Double.isNaN(y)){
           assertTrue(true);
        }else{
           assertEquals(x, y, delta);
      }
   }