issue-202 Fix isnearlyequal floating point comparison

[nicolaslg]

#33 introduces a bug, where isNearlyEqual(a, b) returns false when (a == 0) xor (b == 0) (it is an exclusive or), even when the other argument is near zero.

We might have to properly investigate further for a more reliable check, but this fix should do in the meantime.

Careful, as it affects the Utils::Math::Point operator == it is widely used in all kinds of situations by quite a lot of the methods

[cedricchevalier19]

I am not fully aware of Magix3D coding style, so some of my remarks may not apply.

I think the comparison can be improved using methods like https://randomascii.wordpress.com/2012/02/25/comparing-floating-point-numbers-2012-edition/ or https://realtimecollisiondetection.net/blog/?p=89 .

GoogleTest has also some utilities to do this, see https://github.com/google/googletest/blob/3af834740f58ef56058e6f8a1be51f62bc3a6515/googletest/include/gtest/internal/gtest-internal.h#L336 .

Perhaps you may add a unit test directly calling this specific function with your motivating use case (a == 0.0 )^(b == 0.0)?

[cedricchevalier19]

Ideally, constexpr, but it requires implementing it with only one return.

[nicolaslg]

Thanks for the insightful feedback