Implicit conversion from float -19.837728194726168 to int loses precision