Re: rounding problems

On Tue, May 13, 2008 at 02:36:18PM -0400, Justin wrote:
> Double holds 15 places which is the highest value of precision it can 
> maintain before rounding occurs.
> 
> Is is limit less no, but what is?
> 
> Practically speaking  taking a vale 0.000,000,000,000,001 aka
> 1 trillionth of anything,

But remember that if you add this value onto a large number and then
take off the large number the result will be zero.

  (0.000,000,000,01 + 1,000,000) - 1,000,000  ==>  0
  0.000,000,000,01 + (1,000,000 - 1,000,000)  ==>  0.000,000,000,01

In general, operations on floating point numbers will increase their
errors.

> i view the problem solved for 98% of problems.

Floating point math is good for most problems, hence why most languages
expose the abstraction.


  Sam