ScientificPython is a collection of Python modules that are useful for
scientific applications. Most of them need the Numerical Python
extension (aka NumPy), which is available at http://numpy.scipy.org/.
This is release 2.8 of ScientificPython. All the documentation is in
Doc. To browse the reference manual for all the modules, point your
browser at Doc/Reference/index.html.
Release 2.8 is identical to the development release 2.7.10 except that
the default numerics package is NumPy instead of Numeric. Compared to
previous stable release series 2.6.x, there are numerous new features,
most notable the subpackages Scientific.DistributedComputing and
Scientific.Clustering.
If you find bugs, please tell me, and if you improve something, please
send me the modified version. I don't promise anything, but since I
use these modules a lot for my own work, I have an interest in keeping
them bug-free and usable.
For updates, check
http://dirac.cnrs-orleans.fr/ScientificPython/
from time to time.
Konrad Hinsen
Centre de Biophysique Mol<6F>éculaire (CNRS)
and Synchrotron Soleil
E-Mail: hinsen@cnrs-orleans.fr
---------------------------------------------------------------------------
Installation:
-------------
Required: Python 2.2 or higher. Most modules also require Numerical
Python. If you want to use the netCDF interface module, you also need
the netCDF library, version 3.0 or higher.
1) Unix-like systems, including Linux and MacOS X
Installation is as simple as
python setup.py build
python setup.py install
(the second command requires root privileges on many installations).
This will build the netCDF module if it can find a netCDF installation
in either /usr/local or in /usr. If your netCDF installation is elsewhere,
set the environment variable "NETCDF_PREFIX" to the directory in which
"include/netcdf.h" and "lib/netcdf.so" are located.
2) Windows
To build the netCDF interface module, you must have a binary version
of netCDF. You can either compile it yourself, or download the
precompiled netcdf.dll from the netCDF Web site. You also need the
file netcdf.h, which you can find in the netCDF source code
distribution. PLEASE MAKE SURE THAT NETCDF.DLL AND NETCDF.H ARE FOR
THE SAME NETCDF RELEASE!
Installation is as simple as
python setup.py build --netcdf_prefix=dir_h --netcdf_dll=dir_dll
python setup.py install --netcdf_prefix=dir_h --netcdf_dll=dir_dll
where "dir_h" is the directory in which netcdf.h is located and
"dir_dll" is the the directory in which netcdf.dll is located.
Alternatively, you can create a binary installer using
python setup.py bdist_wininst --netcdf_prefix=dir_h --netcdf_dll=dir_dll
Using numpy or Numeric:
-----------------------
There are three nearly compatible implementations of numeric arrays for
Python: the original one, Numerical Python (module name "Numeric"),
a later rewrite called numarray, and an evolution of Numeric that integrates
features from numarray, called numpy. Numeric and numarray are no longer
supported. It is expected that in the near future, most applications
still using Numeric or numarray will migrate to NumPy.
Starting with release 2.8, Scientific Python uses numpy by default.
However, the support for Numeric and the partial support for numarray
are still there. The choice between one of the three packages is made
at installation time. To use Numeric, install using
python setup.py build --numeric
python setup.py install
To use numarray, install using
python setup.py build --numarray
python setup.py install
Note that the modules
Scientific.Functions.Derivatives
Scientific.Functions.FirstDerivatives
Scientific.Functions.LeastSquares
do not work correctly with numarray because they rely on a feature of
Numeric/numpy that is missing in numarray. For the same reason, the modules
Scientific.Functions.Interpolation
Scientific.Physics.PhysicalQuantities
can be used only partially with numarray, the application of mathematical
functions to the defined objects does not work.
The BSPlib interface:
----------------------
Earlier releases of ScientificPython contained an interface to the
BSPlib library as implemented by the Oxford BSP Toolset. This library
is no longer maintained and cannot be installed easily on most of
today's computers. The BSPlib interface was therefore temporarily
removed from ScientificPython.
There are two other implementations of BSPlib: the PUB library, which
doesn't seem to be maintained any more either, and the BSPonMPI library.
I am investigating the possibility of porting the BSPlib interface to
BSPonMPI, but this will only be in the development releases initially.
Note that the package Scientific.BSP is not concerned by the absence
of BSPlib. It has always had a low-level interface based on MPI, which
is in fact what most users have preferred. This MPI-based interface
continues to work.
---------------------------------------------------------------------------
Examples:
---------
The directory Examples provides a few simple example applications,
including a C extension module that uses the netCDF C-API.
Using Scientific.Visualization.VMD under Windows
------------------------------------------------
If VMD is installed in its default location, everything should work
fine automatically. Otherwise, an error message will be printed. The
default location is
c:\Program Files\University of Illinois\VMD\vmd.exe
or its equivalent on non-English Windows versions ("Program Files"
changes its name).
