[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

About DebianScience wiki

First of all, thanks for the idea of setting up this list. Hopefully it will
be very useful for scientists using Debian.

I was reading the existing wiki at http://wiki.debian.net/?DebianScience.
Although currently the wiki is still only a sketch, I thought of giving a
few comments from my experience about the tools mentioned:

- I think Octave is *really* useful. I don't use all its features, but for
my use, which is basically post-processing data gathered from numerical
simulations. It is also extremely useful when starting the design of
numerical code which might later be developed as C++ code (in my case), to
test ideas and such. And certainly, we have had good experience using it as
a research tool (chaotic maps, neural networks, fluid motion).  

I can't comment on a detailed comparison with Matlab, but someone posted in
the octave list a comment on an unfair perception of Octave during Debconf
(http://www.octave.org/mailing-lists/help-octave/2005/2897). Maybe someone
can discuss that better.

- What I do think is really missing is an adequate plotting tool, either
integrated with Octave or stand-alone. gnuplot output does not look really
good, and fortunately there are many other projects in that direction. But I
haven't found my personal favorite yet. (PlPlot was very, very near, but I
program more fluidly in C++, and that caused me trouble eventually.)
  I also tried R, but I found it difficult to learn for the simple tasks I
intended when I did.
  As I currently only need good looking 2d graphs, I'm using xmgrace. When I
need 3d and colored density plots and so on, I will have to do another round
of research. 

- One advantage of xmgrace is that I can generate .agr files
from data without launching the graphical tool. (A C++ program I wrote takes
the data and creates the .agr file). Since xmgrace can convert those agr
files into eps files from the command line, I can go from the raw data to
the publication quality figure with a single command.

- An alternative to the above is gri, which I tried, and is really nice, but
I prefer ticks inside the frame. gri put the ticks outside, and I could not
find a way to configure it. I tried it in woody though. I admit I haven't
tried after the sarge upgrade. If gri is more configurable, I could try
again. But the point is configurability of the graphical output is a must.
And for me, being able to generate eps figures directly from data from the
command line is very important.

- About mathematical libraries, I would possibly add Matpack. I don't see it
packaged for debian, and that probably is for licensing issues. There was
one particular problem with complex calculus that I had to solve using
Matpack, because GSL didn't implement the special function I needed.

- Another missing link is algebraic calculations. I have tried Maxima, and I
know of several other tools which implement it, either as a complete system
with gui and all that, or as C/C++ libraries, but still Mathematica is the
way to do it. At least for me.

Of course, I understand all the hard work involved in developing these
projects. I only wanted to share my comments. And thanks again for opening
this list.




Reply to: