Implicit conversion from float -22.094594594594582 to int loses precision