Implicit conversion from float -31.312803889789308 to int loses precision