Implicit conversion from float -158.14814814814812 to int loses precision