Playing with Tcl/Tk  (warning: AI)
==================================

Do you remember the Tcl/Tk [1]? It's a programming language (originally
make as  a LISP which  does not  look as a  LISP so it  shouldn't scare
people) and of course an interpreter.  It was very popular in late 1990
and early  2000s because it  have included on  only the "tclsh"  (a Tcl
shell) but also the "wish" (Windowin Shell)  - a nice thing to make GUI
in the  Linux easily (in  reality, it ahs  been ported and  used almost
everywhere  including Windows,  classic and  modern Macintoshes  and so
on). It  also was the  default GUI toolkit for  the Python (I  think it
still is?).

Well, I also  wrote a few simple  GUI tools in the Tcl/Tk  in the past.
Some of  them I still  use or my  (now very infrequent)  research work,
some of them weren't used for decade(s).

Now to the main  point: I do have some older computers  which I use (on
different places  on the Earth). While  SGI O2 can use  a fairly modern
Tcl/Tk 8.4 the SGI Indigo (especially my oldSGI IRIS Indiog R3000 w=chi
runs at 33 MHz  and can use the IRIX 5.3 at best)  stops at Tcl 7.0 and
Tk 4.0 (a version just before  Tcl and Tk numbers became synchronised).
In theory, I can install the Tcl/Tk 8.0 on the Indigo (but I don't have
enough disk space at the moment).

I have found recently the I still have a source code for a Tcl/Tk based
tool for structural analysis (well,  just for computing of deformations
and internal  forces, it's  called fem1.tcl for  some reasons).  It was
originally meant as an  interface form my C code but  I never find time
to make  them cooperate. Moreover, I  revived the project in  2020 (the
COVID time) and  added a basic computing kernel (called  test.tcl as it
was generally meant just for testing..)  amd also I also added plotting
of deformations and internal forces  (very, very basic). From that time
I never returned to finish it.

Anyway, I  was in need of  such tool as commercial  products don't work
well  on the  ppc64le (there  are  none, actually)  and available  open
source tools (including  my own, unfortunately) are too  complex or too
limited for my use.

There  were recently  some AI-related  courses  at my  university so  I
attended them  and decided  to try to  use my new  knowledge to  try to
improve my  code (and to see  how useful or useless  these things are).
For practical reasons ($$$ - the university is already paying for this)
I used the gemini-cli tool to interface the Google's product.

Well, it's a bit  crazy. It helped me to find a lot  of old bugs but of
course it introduced their own.  But by iterative procedure (well, it's
try-test-find errors-repair errors) I have  been getting a quite useful
piece of code (at least for me).

It  seems that  starting  from  actually existing  and  (less or  more)
working and complete code was a good  idea. I also tried to order it to
generate  (a much  simpler tool  from nothing).  It worked  (after some
iterations)  and it  looked much  better visually  than my  code but...
(well,  it almost  works now,  I just  need to  resort some  thing like
non-working EPS/PS export  and so on -  but this tool is  very tiny and
basic; this  is not  an issue  as my  more complex  tool is  Tcl/Tk 8.0
compatible so  I still  need a simple  tool - 7/4  compatible -  for my
slowest Indigo).

The conclusion is that's it's slow and painful process, but if I have a
solid base then  improvements are much faster  than if I try  to do the
same with any help. At least for something is that AI thing useful...


The good  thing is that  I have had  to refresh my  Tcl/Tk progrramming
knowledge. It's a strange language but it works. :-)

P.S. Thinking where  I should place the  sources as the SDF  Git is not
publict (and it's the  only Git server I still can  reach from the IRIX
machine). An olds-school tarball on my WWW page maybe? [2]


Written on the SGI O2, of course.


References:

[1] https://tcl.tk
[2] http://jirka.sdf.org/fem1.tgz