Implicit conversion from float -22.17228464419476 to int loses precision