Implicit conversion from float -0.14117647058822058 to int loses precision