How to compare floating point variables in C++?

Question

This post stated that you shouldn't compare floating point variables with == because of rounding errors. What should I use then, and when?

score 3 · Answer 1 · answered Apr 04 '16 at 12:20

3

You could use something along the lines of if (abs(result - expected) < 0.00001)

or for a relative, rather than absolute, error:

float relativeError = abs((A - B) / B);
    if (relativeError <= maxRelativeError)

answered Apr 04 '16 at 12:20

Pedro

But beware. That "relativeError" calculation blows up if B is zero. In general, if you use relative errors, zero will only ever compare "nearly equal" to a value which is exactly zero. – Martin Bonner supports Monica Apr 04 '16 at 12:33

1 Answers1