[Pythoncard-users] FW: scipy

[Kevin A. <al...@se...>]

I discovered SciPy http://www.scipy.org/ two days ago. Originally, I was
interested in graphic format conversions, such as numerical arrays (NumPy)
that we might need to support in PythonCard similar to the support I added
for PIL. I still need more feedback on what should be put in the PythonCard
framework for PIL and NumPy.

Meanwhile, I've been in contact with Eric Jones of the SciPy team and have
installed SciPy and run through the tutorials at:
http://www.scipy.org/site_content/tutorials/plot_tutorial

Since the plt module of SciPy is based on wxPython you can drive plt from
the PythonCard shell (PyCrust) without needing the line:

   >>> import gui_thread

gnuplot also seems to work fine, but I didn't use it for very long so there
may be some window manager conflicts between gnuplot and wxPython. As a
simple exercise I modified minimal.py to simplify data input, a screen shot
is attached. This will probably be expanded and become another sample for
PythonCard; the sample will require SciPy which includes/requires NumPy. The
short story is that PythonCard and SciPy seem like a good match. The
question becomes what level of integration and overlap should we have
between the two projects.

I don't know much about the requirements for "scientific computing" so right
now I'm at the stage of asking newbie questions about what scientific users
need in a GUI framework and environment. I thought Eric's response below was
worth forwarding to the list in case other people have suggestions on how
PythonCard can be a useful tool in science.

If you're on the PythonCard list are interested in this topic, please speak
up.

ka

-----Original Message-----
From: eric jones [mailto:er...@en...]
Sent: Thursday, October 11, 2001 10:53 AM
To: Kevin Altis
Subject: Re: some other questions on scipy


> I noticed SciGraphica today on the SF Python list.
> http://sourceforge.net/projects/scigraphica/
>
> It only works on Linux, but I guess this is a "competitor" of sorts?

Not really a competitor -- SciPy has a huge amount of Numerical calculation
stuff that is orthogonal to SciGraphica's effort.  They are really about
making it easy to build pretty plots.  I'm very in that project as they look
to have taken the graphics a very long way.  Further, they are planning on
breaking things out to make them accesible from Python command line in the
next version which means they can be used right along with SciPy.

The big problem with SciGraphica is its based on GTK.  GTK looks bad on
Windows, isn't easy to install, and isn't stable.  Windows is a major
platform and SciPy has to support it as well as, or even better than Linux,
to be viable.  After all, Linux people are used to fiddling with things to
get them to work.  Windows people aren't.  That means GTK isn't a viable
option in my view.

On the other hand, I have wondered how hard it is to switch this general
framework over to wxPython or even try to set it up so it is backend
independent.  I have been talking with the guys at the Space Telescope
Institute (they run Hubble), and they are also very interested in graphics
stuff.  We have pitched around this idea quite a bit in the last few weeks,
but haven't formalized a set of specs or a game plan yet with regards to
this.

We are also looking hard at using VTK (www.kitware.com) for all 3D stuff.
There are benefits and drawbacks to this that are still under consideration.

> What are other comparable packages, anything driven by Python or that
Python
> interfaces to? I'm just trying to understand the bounds of the problem
that
> needs to be addressed.

I don't think SciPy really has a competitor.  The Scientific Python project
by Konrad Hinsen is similar, but is slightly more focused on computational
chemistry.  The most similar package to
SciPy would be Matlab (or Octave in the Open Source version).  National
Instruments has a less used one called HiQ.  There is also IDL, Mathematica,
Maple, etc. which serve a similar purpose.  The general goal of SciPy is to
make python a viable alternative to these tools and then leverage the power
of Python to take it beyond them.  If you have access to Matlab, you might
try it out.  Octave is less useful, but still a reasonable thing to look at.
I've also heard of a language called R that is OS that might serve similar
purposes.

So SciPy proper I guess focuses on two things: scientific libraries and
visualizing data.  We want to make it easy to do scientific stuff in Python
as well as make sure that the tools or algorithms/scripts developed with
SciPy can be dropped into larger apps with minimal fuss.  Things like turtle
might  actually have a place in the visualization section.  The GUI building
end, at least right now, is stuff we'd like to borrow from you guys that
know what is going on in that arena of Python.  Later, we might have a full
blown GUI with all this stuff available together though.  I'm very glad your
playing with SciPy because combining SciPy with PythonCard is *exactly* what
is so powerful about building it in Python -- and I'm counting on you et al.
to think of things we never had. :)

> I can't do the sample with !NewAxis:

This is a mistake.  Thanks for finding it.  Where is it and I'll fix it.

see ya,
eric