Implicit conversion from float -7.474747474747474 to int loses precision