🚀 go-pugleaf

In article <BtKpKz.M3F@netnews.jhuapl.edu>
pwm@netnews.jhuapl.edu (Miller Phil F1E x8335 ) writes:

    Also I would like some advice on the best way to learn
   C++ and object oriented programming with numerical
   simulation in mind.

I see that other posters have recommended books for learning C++, so I
won't repeat them here.

I wanted to tell you, however, that I've been using C++ to design an
object-oriented approach to constrained optimziation and it's been working
beautifully. I can write a class to exploit the special structure of a
particular constraint function, and my optimization engine doesn't have to
change one bit, because the internal stucture of the constraint is hidden
from it. On the other hand, I can write different classes to exploit
different optimization techniques, and similarly the constraint code does
not need to change.

My code is extremely modular, easy to maintain, and optimized for speed
all at the same time (if I do say so myself :-)

Of course, object-oriented programming is not a complete solution to
numerical programming. As a matter of fact, I use the FORTRAN LAPACK
functions for my matrix algebra, because it suits my particular
application. There are some excellent matrix classes out there, though, as
others have recommended. I have tried out M++ from Dyad, for example, and
it's quite nice.

Mike

>>C++ and object oriented programming with numerical
>>simulation in mind.

You should learn the language for what it is before trying
to bend it to do what you want.  I'ld suggest leafing through
"C++ Primer" by Stan Lippman (get the 2nd edition), and then
write yourself a few useful non-numerical programs.  Then
read Chapters 11, 12, 13 of "The C++ programming language"
by Bjarne Stroustrup (2nd edition).  

Then start solving the problems you really want to solve.  After your
first prototype, leaf through "Advanced C++" by James Coplien to
remind yourself that you still have a lot to learn :-)

-mark

In article <BtKpKz.M3F@netnews.jhuapl.edu> pwm@netnews.jhuapl.edu (Miller Phil F1E x8335 ) writes:
.
.
.
>
> Also I would like some advice on the best way to learn
>C++ and object oriented programming with numerical
>simulation in mind.
>

I would recommend starting with Bjarne Stroustrup's _The C++
Programming Language_, 2nd Edition (Addison-Wesley, ISBN
0-201-53992-6, don't waste money on 1st edition).  In my opinion
there are no books that can help with numeric programming in
particular in C++.  S. Lippman's _C++ Primer_ is also good
(Addison-Wesley, ISBN 0-201-16487-6).

The next best thing would be to get some matrix classes.  The best are
those included in the M++ class library.  It is a very extensive set
of classes that essentially turns C++ into a full-featured array
language.  In fact if you've had some experience in array languages
you would be able to start programming at once. It includes LINPACK 
and EISPACK functions.  For more information write, call, or fax

    Dyad Software
    515 - 116th Ave NE, Suite 120
    Bellevue, WA 98004
    (206) 637-9427 (voice)
    (206) 637-9428 (fax)

I would be happy to send anyone some sample M++ code on request.

RogueWave distributes a commercial product that provides a set of matrix and
factorization classes more limited than than M++'s classes.  For more 
information write, call, or fax

   Rogue Wave
   P.O. Box 2328
   Corvallis, OR 97339
   (503) 754-2311 (voice)
   (503) 757-6650

Rogue Wave also distributes a free booklet titled "How C++ Can Make a Better 
FORTRAN".

newmat04 is a public domain class library.  It doesn't have the sophisticated 
memory management of M++, nor the LINPACK/EISPACK functions, but it is free.
This class library would be very usefull for getting started in C++.  It can
be retrieved by anonymous ftp from ftp.uu.net.  It is in the directory
usenet/comp.sources.misc/volume26