Judging floating point numbers cannot use constant equal sign about jsbsim HOT 8 CLOSED

Last call before departure closing the issue as won't fix 😃

from jsbsim.

I'm assuming he means the method should be along the following lines:

bool operator==(const FGColumnVector3& b) const {
   return fabs(data[0] - b.data[0]) < 1e-13 && fabs(data[1] - b.data[1]) < 1e-13 && 
      fabs(data[2] - b.data[2]) < 1e-13;
}

from jsbsim.

I'm not sure I'm getting your point: your statement is too vague. Can you please point at a place in the code where you think the usage of == is "not right" ?

from jsbsim.

Ping !

@jinzhu1911 It's been one month since I asked you some clarifications but got no answer. Could you please be more specific ?

from jsbsim.

In the function Comparison operator.

 bool operator==(const FGColumnVector3& b) const {
    return data[0] == b.data[0] && data[1] == b.data[1] && data[2] == b.data[2];
  }

from jsbsim.

@jinzhu1911 Your answer isn't clear. Could you elaborate?
Please, do keep in mind that this is not the place to be cryptic.

from jsbsim.

in the function Comparison operator.
bool operator==(const FGColumnVector3& b) const {
return data[0] == b.data[0] && data[1] == b.data[1] && data[2] == b.data[2];
}

I'm assuming he means the method should be along the following lines:

bool operator==(const FGColumnVector3& b) const {
   return fabs(data[0] - b.data[0]) < 1e-13 && fabs(data[1] - b.data[1]) < 1e-13 && 
      fabs(data[2] - b.data[2]) < 1e-13;
}

Yes, I mean this. Thanks very much!

from jsbsim.

I'm assuming he means the method should be along the following lines:

bool operator==(const FGColumnVector3& b) const {
   return fabs(data[0] - b.data[0]) < 1e-13 && fabs(data[1] - b.data[1]) < 1e-13 && 
      fabs(data[2] - b.data[2]) < 1e-13;
}

Yes, I mean this. Thanks very much!

Well FGColumnVector3::operator== tests equality of two FGColumnVector3 instances which is different from testing if they are almost equal. Test frameworks often provide the latter functionality but in that case, they let the user specify the precision with which the values are compared. C++ does not allow to specify an additional parameter to operator==.

If that really is a problem, I'd rather add a function such as

bool AlmostEqual(const FGColumnVector3& a, const FGColumnVector3& b, double precision) {
  return fabs(a[0]-b[0]) < precision && fabs(a[1]-b[1]) < precision && fabs(a[2]-b[2]) < precision;
}

and leave FGColumnVector3::operator== as it is.

from jsbsim.