🚀 go-pugleaf

In article <1992Sep22.103137.7811@gdr.bath.ac.uk> masmm@gdr.bath.ac.uk 
(M McGettrick) writes:
> 
>               It seems to us that different software systems are being used 
> to solve different aspects of problems in materials science.  Recently some
> "integrated" packages have appeared which can tackle all aspects of a problem
> from numerical analysis to computer algebra to visualization.
>               In connection with some work we are doing on linking symbolic
> and numerical computer packages, we would like to know
> 
> a) the requirements (if any!) materials scientists have for integrated 
>    (software) systems,
> 
> b) what possible applications have such systems in materials science,
> 
> c) whether the packages available at the moment are used widely and whether
>    they satisfy current needs.
> 

You do not give any examples of the "integrated packages" you have in mind,  
perhaps because you feel the examples are obvious?   Macsyma, Mathematica,  
Maple, IDL, Theorist, Matlab?   Maybe you have a "Rorschach test" in mind, and  
are interested in what respondents call an integrated system?

Anyway...
Re (a).  My requirements include:
*  I have to be able to learn the system (or the part of it that applies to the  
problem at hand) faster than I con solve the task using existing methods.
*  There should be examples of the use of the system on problems similar enough  
to mine so that I can convert an example to my particular problem.
*  The results have to be right.  (It helps to have test cases available, with  
automatic methods to run the system on the test data, and compare the results  
with validated answers.  It also helps to have the source code available, and  
the algorithms, too.)
*  Running the package should not crash my computer, or destroy disk files.
*  This results have to be exportable to other software packages.  Text should  
be convertible to ascii (or to nroff-format or to TeX-format).  Two-D graphics  
should be in eps-format or exportable to it.  Three-D graphics should be in a  
:standard" schema.

Re (b).  My applications...
I use Mathematica and Maple (M&M) on a daily basis --- I find that these  
products are excellent for closely interactive work on "small" problems, and  
have changed my life for the better.  I now engage with enthusiasm jobs that I  
dreaded before, such as plotting a set of (say) 10,000 (x,y) pairs in various  
ways until I find the way that clarifies a relation, or effecting a non-linear  
least square fit to a data-set (Thanks to Pekka Janhunen!), or solving a set of  
several non-linear symbolic equations, or evaluating an integral, or  
establishing the first several terms of a Taylor's series of a complicated  
expression, or finding a Pade approximation, or using an implicit Runge-Kutta  
scheme to numerically integrate a stiff set of six coupled non-linear  
equations, ... .  (Each of these is a task I carried out in the last week.)

"Small" is any problem that turns around in less than a few seconds, or maybe  
several minutes at most.  For "large" jobs, I often extract a "representative"  
part and explore with M&M until I find a good viewpoint.  Then I code the task  
using C and the library from Software Tools (Press, et al.) and run the full  
job.  So M&M are almost invariably helpful, even if the final work is carried  
out more conventionally.

Re (c):  Used widely?  Gosh, I can speak only for what I see going on in my  
lab, and on the floor around me.  About half of us are using "integrated  
tools".

Satisfy current needs?  I do not find M&M to be the right tools for producing a  
*complete* graph, or for writing a report.  And I am of the opinion that these  
powers are inappropriate to add to M&M --- it would make them rather larger,  
and more prone to failure, and would slow down development of their  
mathematical powers --- but it is just for their mathematical powers that I use  
them.

I think people use "integrate" packages for mainly two reasons:
(1) different programs tend to have very different user-interfaces, so that  
there is an advantage to learning just one program (if it can actually do  
everything you want), and
(2) different programs tend to use very different internal data-structures, so  
that it can be difficult to use the data produce by one program within another  
program.  Hence, there is an advantage to doing all of one's work within one  
program.

But neither of these reasons reflects 
          --> the way it has to be <--
only
          --> the way it is (on some systems) <--.

And this is *not* this way on the NeXT computer.  I routinely copy text and  
graphics out of M&M and into drawing programs and into text programs.

** It is the computer itself (hardware and operating system) that can and  
should provide the integration. **

All the software programs have to do is allow the operating system to provide  
the integration.

I am happy with TeX (actually, LaTeX) as a formatting language for producing  
reports and memos.  I doubt that anything much smaller (and less complicated)  
than TeX/LaTeX could produce high-quality results.  I doubt that Mathematica or  
Maple would benefit much from having a TeX-class system incorporated within  
them --- nor need they!  All that is necessary is that they communicate  
gracefully with a TeX-system.

Similarly, I am happy with encapsulated PostScript (EPS) as a language for  
producing graphics.  I find that I want something with at least the power of  
Adobe Illustrator (and Create and Diagram, two drawing programs for the NeXT  
computer) as a way of working with EPS.  I doubt that Mathematica or Maple  
would benefit much from having such a set of tools added.  Nor is it necessary  
to add them.  All that is necessary is that Mathematica and Maple prepare their  
graphics in a way that can be further manipulated by a top-rate drawing  
program.


Regards,
Henning Leidecker