[PYTHON MATRIX-SIG] NASA Research Announcement - Astrophysics Data Program

Jeffrey Templon templon@studbolt.physast.uga.edu
Thu, 6 Mar 1997 15:49:03 -0500


Paul Barrett writes:

 > 1) To make Python more accessible to the scientific community,
 > Numerical Python needs to be provided as an easily installable
 > package, preferably having pre-built binaries for common platforms
 > available.  A binary package will make it much more likely that a
 > scientist will install the package to evaluate it.

I fully agree.  Asking your average Joe Physicist to go thru the
Makefile procedure will never work.  If you give EXPLICIT instructions
(both in 'sh' and 'csh' language, for Unix people) on how to set the
environment variables, you may succeed.

 > 2) A basic (or standard) data analysis environment must be developed.
 > Modules to be included in this package would almost certainly include
 > the numeric module, the netCDF module, the GIST module, and the PIL
 > module which would provide basic array, file, plotting and graphic
 > capabilities, resp.  The basic numerical python package would allow
 > most scientists to quickly and easily plot or display their data.
 > Note that most of these modules are already being developed and only
 > need to be integrated into a standard data analysis environment.

This is in my mind the most important thing, and also would be a HUGE
contribution to the NumPy community.  I used GIST about a year ago,
and while it was clear that it was nice, it was a hassle to use.
There were conflicts between it and the standard GNU Readline
mechanism, for example, and also lots of little functions and
parameters to remember.  One would hope for some default function
which would look sort of like GNUplot (or at least is as simple):

	plot1=gist.mkplot()       # default plot: 2D lin-lin


	plot2.xlabel('Energy (MeV)')

I guess I'm trying to say that GIST plotting was very 'unpythonish' a
year ago.  I haven't looked since (too busy with other stuff) so I
apologize if these remarks are now off base.

The bottom line: modules listed in your (2) which have a consistent,
Pythonnish syntax and form, would make Python (and NumPy) much more
attractive to the community at large.  Go for it!


MATRIX-SIG  - SIG on Matrix Math for Python

send messages to: matrix-sig@python.org
administrivia to: matrix-sig-request@python.org