Andy Armstrong andy at
Tue Jan 16 22:37:11 GMT 2007

On 16 Jan 2007, at 22:16, Paul LeoNerd Evans wrote:
>> On 15 Jan 2007, at 17:14, Peter Hickman wrote:
>>> For the sake of argument X is imprecise and 0 is precise. Just how
>>> many such cases exist for the various operations I do not know. But
>>> doing this in software is not going to be cheap. Quite what you
>>> would gain in knowing that a number was imprecise I do not know,
>>> but them I'm a web developer :) not an accountant.
>> And PI * anything-other-than-zero is always going to be a big old
>> result if you want to capture all of the bits :)
> If you want to do this sort of thing properly, you keep it symbolic  
> and
> abstract within the logic, and only try to convert it to decimal  
> when you
> print it. That way, if you want to print, say, 10 decimal places
> accurately, you lazily evaluate with however many digits were  
> needed in
> the first place to guarantee that.

What happens if you just want to do lots of calculations with the  
value of pi or e or some other irrational number? At what point does  
the value get consummated?

Andy Armstrong,

More information about the mailing list