Implicit conversion from float -2.2837370242214377 to int loses precision