Implicit conversion from float -0.3579952267303099 to int loses precision