Implicit conversion from float -0.19867549668873608 to int loses precision