code not true?

[Gordon McMillan]

Michal Wallace wrote:

> On Sun, 23 Apr 2000, Kragen Sitaker wrote:
> > That said, it *is* just a speed issue --- there are representations
> > other than floating point, like the ordered-pair rational-number
> > representation, that give the right answer but are slower to work
> > with.
> 
> But... Python is a scripting language.. It's okay to be slow.  

It's not just a scripting language. Because numbers are 
objects (not primitives), Python is already slow in doing math, 
and the cumulative complaints about that are voluminous.

> From
> what I'm hearing, many people on this list shy away from floats simply
> because of the innacuracy.. Why don't we use the way that gets the
> right answer?

Some tech magazine columnist observed that there are 3 
types of float users: 
 1) those who think it's easy, 'cause they're just like ints
 2) those who think it's hard, 'cause they're nothing like ints
 3) those who really know.

We've got exactly one 3 (and he's busy - having just bought a 
CDROM with 660M of one-time pad <wink>). The transition 
from 1 to 2 is hard (as witnessed by this thread). But once 
you're a 2, it's not that hard to learn to use floats sanely.

The theoretically correct alternative is rationals. But they're 
slow and use unpredictable amounts of memory. I think 
Python should have BCD as an alternative. While slow, they're 
predictable in memory use, and the mathematical 
shortcomings they have are exactly those that a naive user 
expects.

But I, despite being in category 2 for close to 20 years, 
definitely don't want to be without floats.



- Gordon