# [PD] [OT] "processing" application

Mathieu Bouchard matju at artengine.ca
Sat Oct 14 13:24:14 CEST 2006

```On Fri, 13 Oct 2006, David Powers wrote:

> When working with computers, I think you'd also want to point out that a
> 'float' has some amount of precision, and that this is related to how
> many bits of memory the computer uses to store the number.

yes

> I believe it would be correct to say a float is a finite decimal
> approximation

No, it's a binary approximation, which is not equivalent to a decimal
approximation because the bases do not have the same prime factors. In
that sense, base 8 and 16 are equivalent to each other, and bases 10 and
20 and 50 are equivalent, but the two groups of bases are not, because one
is 2^a and the other is 2^a * 5^b. (a,b must be more than zero of course,
else they'd be the same for b=0).

> Anyway, knowing that the number is a decimal approximation might be
> important from time to time in multimedia programming...

Knowing that it's not a decimal approximation might be important, at least
for the fact that most of the time 1 - 0.9 is not exactly 0.1, when you
ask pd or most any other piece of software.

You may also use true rationals in some languages, so that 1 - 9/10 is
always exactly 1/10, but it's a lot slower, and it's not available in pd
yet. (It's available in Lisp, Smalltalk, Ruby, Mathematica, and a bunch
more)

_ _ __ ___ _____ ________ _____________ _____________________ ...
| Mathieu Bouchard - tél:+1.514.383.3801 - http://artengine.ca/matju
| Freelance Digital Arts Engineer, Montréal QC Canada
```