The commitee has accepted some version of the following classes
in the new standard:

bits<N>   - template class for fixed-sized sets of bits
bitstring - sets of bits for which the size may be changed
complex   - complex numbers
dynarray  - general array class
string    - simple character string class
wstring   - wide character string
xalloc    - exception for memory allocation errors

(Some vendors, including Borland, already ship some of these
 classes with their C++ compilers.)

The commitee has also been looking at adding some type of
classes for numerical and scientific computation, e.g., vector
and matrix classes. Does anyone know of the latest word on
these?

Are there any preliminary examples or documentation on
the numerical classes the committee is considering? What are the major
issues they are debating?

Kevin Copps
kc@isc.tamu.edu
Institute for Scientific Computation
Texas A&M University

    Kevin Copps writes:

>The commitee has also been looking at adding some type of
>classes for numerical and scientific computation, e.g., vector
>and matrix classes. Does anyone know of the latest word on
>these?

>Are there any preliminary examples or documentation on
>the numerical classes the committee is considering? What are the major
>issues they are debating?

    I don't know what the committee has been considering, but here's something
that they should consider about vector and matrix classes.  The LAPack group,
led by Jack Dongarra, is developing LAPack++, a C++ implementation of the
ubiquitous Fortran LAPack.  Version 0.9 of LAPack++ is now available from
netlib.att.com in pub/netlib/c++/lapack++.

    If X3J16 wants to make numeric classes part of the standard C++ libraries,
and they should, LAPack++ should be given **serious consideration by those who
know and care about numeric computing**.  (This leaves Stroustrup out.)

Dean Schulze