For even more amusement try comparing NaNs ...

--
Tim Ward
Brett Ward Limited - www.brettward.co.uk

   I believe IEEE's floating point standard defines 0.0 as being equal
to -0.0. Probably because it would generate the least astonishment in the
developer.

   - Oliver

gk schrieb:

Hi!
What have you expected?

Of course -0.0 is the same as 0.0.
If you try System.out.println(0.0==-0.0); it will print true.

Because it is true - and mathematics says that there is only one neutral
element for addition and that is 0. And whether you call it 0.0 or -0.0
doesnt matter, because it still is the same.

bye, Ralf

The core problem here is that the IEEE 754 zeros have two functions:

1. A representation of true zero, the additive identity element. A true
zero is less than all positive numbers, greater than all negative
numbers, and does not itself have a meaningful sign.

2. A way of representing incredibly small numbers. It is useful, in some
algorithms, to preserve the sign even when a number is too small to
represent any part of the magnitude. It makes the difference between 1/x
being positive or negative infinity. It can allow some code to work
without explicit overflow or underflow checking.

However, it does lead to a slightly split personality. To make the true
zero behavior work right, it is essential to treat both zeros as being
equal. To make the sign-preserving underflow behavior work, the min of
the two zeros is the negative one etc.

Patricia