What is "-0.0"

Thread view:

Morgan Cheng - 13 Oct 2007 13:06 GMT

In java 6.0, I have below code.
double x = 0.0;
double y = -10.0;
double z = x / y;
System.out.println("z = " +z);

The result is "-0.0". This is wired, why it is negative zero?

Reply to this Message

Richard Reynolds - 13 Oct 2007 15:18 GMT

> In java 6.0, I have below code.
> double x = 0.0;
[quoted text clipped - 3 lines]
>
> The result is "-0.0". This is wired, why it is negative zero?

http://en.wikipedia.org/wiki/%E2%88%920_(number)

Reply to this Message

Patricia Shanahan - 13 Oct 2007 15:49 GMT

> In java 6.0, I have below code.
> double x = 0.0;
[quoted text clipped - 3 lines]
>
> The result is "-0.0". This is wired, why it is negative zero?

I agree that it is weird, but there are reasons why IEEE 754 floating
point arithmetic distinguishes positive and negative zero.

A zero may be a stand-in for a number of very small absolute magnitude.
There are some algorithms where intermediate results overflow and
underflow, but the final answer is representable.

Preserving the sign on underflow, and through arithmetic using a zero,
also preserves the sign on division by the result of the underflow,
leading to the correct choice between positive and negative infinity.

Patricia

Reply to this Message

Stefan Ram - 13 Oct 2007 17:29 GMT

>The result is "-0.0". This is wired, why it is negative zero?

The first result for

http://google.to/search?q=negative+zero

currently is

http://en.wikipedia.org/wiki/%E2%88%920_(number)

.

Reply to this Message

Roedy Green - 14 Oct 2007 10:20 GMT

On Sat, 13 Oct 2007 05:06:35 -0700, Morgan Cheng
<morgan.chengmo@gmail.com> wrote, quoted or indirectly quoted someone
who said :

>The result is "-0.0". This is wired, why it is negative zero?

Because floating point has both a positive and negative 0, signed
magnitude, unlike ints. see http://mindprod.com/jgloss/ieee754.html