Strange math behaviour... could someone explain this ?

[Mark Wieder]

Alex-

Friday, October 7, 2005, 11:14:47 AM, you wrote:

> It's NOT a Rev bug - it's an artifact of double precision binary
> arithmetic (or, if you like, an artifact of the IEEE format used by 
> Intel (and everyone else)).

I'm sure if I coded the same thing in java I would also come up with
179. It's an artifact of trying to force an analog world into a
digital frame and back again. You can see the same sort of thing at
work if you look into "random" numbers.

However, 36-34.2 = 179 is just WRONG.

While it's easy to rationalize the error by examining the inaccuracies
due to the constraints forced by the math libraries, the real-world
consequences become more apparent when the discrepancy is a larger
percentage of the result.

Try this:

on mouseUp
  local t1
  
  put 36 - 34.2 into t1
  add .2 to t1
  
  put t1 && "+" && t1 && "=" && trunc(t1+t1)
end mouseUp

I'd be happiest if the trunc() function internally appended a trailing
"" before performing its task, and I think it's worth a BZ enhancement
request entry - jbv, want to do the honors?

-- 
-Mark Wieder
 mwieder at ahsoftware.net