Right - if you have a monadic function that takes a 32-bit float, your tests should probably literally cover every single input value.