15 Apr
2010
15 Apr
'10
12:02 p.m.
On Apr 15, 2010, at 12:53 AM, roconnor@theorem.ca wrote:
I'd call them disrespectful functions, or maybe nowadays I might call them improper functions. The "good" functions are respectful functions or proper functions.
There's no need to put these into a different class. The IEEE defined this behavior in 1985, in order to help with rounding error. Floats and doubles are NOT a field, let alone an ordered field. 0.0 =/= -0.0 by design, for floats and doubles. 0.0 == -0.0 for integers, exact computable reals, etc. The problem isn't the functions, or the Eq instance. It's the semantics of the underlying data type -- or equivalently, expecting that floats and doubles form an ordered field.