Boost logo

Boost :

From: Marsh J. Ray (marsh.boost_at_[hidden])
Date: 2006-06-04 19:23:55


Anyone else feel that there might be a better name out there for this
facility than "Infinite Precision Integer"?

Taking the first sentence of the draft:
> The infinite precision integer is an integer that is not limited in
> range by the computer word
> length.
I'm certainly no expert on math terminology, so someone please help me
out if I'm confused here, but aren't "precision" and "range" quite
distinct concepts?

It seems like "precision" isn't really what distinguishes this facility
from that of the built-in integral types. E.g,. in what way can't the
built-in int already represent its values with "infinite precision"?
Ref: http://www.open-std.org/JTC1/SC22/WG11/docs/iso11404.pdf "ISO/IEC
11404:1996 Language Independent Datatypes (LID)" Defines 'Integer' as
"the mathematical datatype comprising the exact integral values" with
the properties of "exact" and "unbounded".

Also, I'd stay away from the term infinite, especially when it's clearly
constrained by something as finite as computer memory. Terms for "the
infinities" are already defined by IEEE-754 have a meaning different
from what seems to be intended here.

Perhaps a better name for this facility might be "Unbounded Integer" or
"Unrestricted Range Integer". (De-ranged Integer anyone?)

In keeping with the C/C++ traditions of referring to the
representational size (short, int, long, long long, etc.), another
approach to might refer to them as "Unsized" integers.

I can see why other libraries are called simply "bignum", "gmp", etc..

- Marsh


Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk