[racket] Multiplying by 0

On Feb 14, 2011, at 3:26 PM, David Van Horn wrote:

> On 2/14/11 3:14 PM, Sam Tobin-Hochstadt wrote:
>> No, it's not a bug.  Since 1e100 is an inexact number, there's
>> uncertainty about the minimum of those two numbers, and the result is
>> therefore inexact.
> 
> I would've expected min to return a number that is `eq?' to one of its arguments.
> 
> In other words, what is wrong with this definition of `min'?
> 
> (define (min n1 n2)
>  (cond [(<= n1 n2) n1]
>        [else n2]))

What's wrong with this is that, mathematically, since 1e100 is inexact, we're not CERTAIN it's >= 0, so the "proper" answer to (<= n1 n2) is not true but rather almost-certainly-true.  (An "inexact Boolean", if you will....)

When you define the function as above, the "<=" takes its best guess as to which number is really smaller and pretends that the answer is certain.

Stephen Bloch
sbloch at adelphi.edu