Re: Bug with numeric calculation

[Chuck Swiger]

On Aug 17, 2006, at 1:10 AM, Dr. C. Griewatsch wrote:
> echo 1.0 1.1 | gawk '{print 0.1+$1-$2}'
> 0
>
> echo 1.1 1.2 | gawk '{print 0.1+$1-$2}'
> 2.22045e-16
>
> The same error can be reproduced with gawk 3.1.5 on the same  
> machine, or
> with gawk 3.1.0 on Solaris 5.8, and with gawk 3.1.4 on Suse 9.3.

Floating point numbers are represented at finite-precision binary  
values, which is discussed in considerable detail here:

  http://en.wikipedia.org/wiki/Floating_point#Problems_with_floating- 
point

--
-Chuck