From: Helmut Zeisel (helmut.zeisel_at_[hidden])
Date: 2001-06-20 23:39:39
20 Jun 2001 David B. Held wrote:
> >Is there any interest in an unlimited integer class?
> >I wrote an unlimited integer class
> >using the classical O(n**2) algorithms
> >for multiplication and division,
> >so it works reasonable fast up to some 100 decimal digits.
> I wonder if it would be even more generally useful if it were
> generalized to a fixed-point representation? Just an idea.
IMHO an unlimited integer class might be
a good starting point for several generalizations,
such as the already existing boost::rational
or some fixed-point or high precision floating point classes.
I am not sure, however, whether they should be directly implemented
in the integer class or whether an independent adapter
(such as boost::rational) could do the same job.
> Did you implement it in a binary representation or something
> like BCD?
The radix (and its type) is a template parameter,
so the radix can portably be everything from 2 to 65536
(and perhaps 2**32 on a 64 bit machine).
The boolean operators &,|, and ^, however,
work only corresponding to our conventional logic
when the radix is a power of 2 ;-)
Using char digits with radix 10, you can have a kind of BCD.
You have, however, only one decimal digit per byte,
not two digit as in real BCD.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk