I think I’d have a different implementation for floats that would return the nearest two it’s between, and punt the tolerance decision up a level to the calling code.
I think I’d have a different implementation for floats that would return the nearest two it’s between, and punt the tolerance decision up a level to the calling code.