Implicit conversion from float -78.01566579634465 to int loses precision