Python

We provide as many modules as possible that come with the current Debian GNU/Linux stable release. Nevertheless, that might not be enough for your needs since you may want to use the newest version of some module or one that is not part of Debian.

Since Python 2.6 there is an easy way to install missing or outdated modules in your home through easy_install.

How to use easy_install

• Command line help:	easy_install --help
  Online documentation:	http://packages.python.org/distribute/easy_install.html
  Install a new module:	easy_install --user MODULENAME
  Update an existing module:	easy_install --user -U MODULENAME

Modules will be installed in your home within ~/.local/. You do not need to adapt the PYTHONPATH environment variable since python will look for modules in this directory automatically.

Installing other versions of Python

You can of course install other versions of Python in your home. A very comfortable way of doing that is by using pythonbrew. You will find a howto on that website with detailled instructions how to use it.

Installation of custom (non easy_install-able) Python modules in the home directory of a user

We are sometimes asked for newer version of Python modules. Since we do not want to maintain Python for Linux in SEPP, the user needs to install these modules in his/her home directory.

On this page we will list some bash-snippets that install some often requested modules in a users home.

numpy

VERSION_NUMPY=1.6.0
installdir="${HOME}/opt"
builddir="/scratch/${USER}/build/numpy"

export PYTHONPATH=${installdir}/lib/python

mkdir -p ${builddir}

cd ${builddir}
wget --output-document=numpy-${VERSION_NUMPY}.tar.gz \
        http://sourceforge.net/projects/numpy/files/NumPy/${VERSION_NUMPY}/numpy-${VERSION_NUMPY}.tar.gz/download
tar -xvvzkf numpy-${VERSION_NUMPY}.tar.gz
cd numpy-${VERSION_NUMPY}
python setup.py build --fcompiler=gnu95
python setup.py install --home=${installdir}

scipy

• First you need to install scipy as shown above and make sure PYTHONPATH points to the new numpy installation.

VERSION_SCIPY=1.6.0
installdir="${HOME}/opt"
builddir="/scratch/${USER}/build/scipy"

export PYTHONPATH=${installdir}/lib/python

mkdir -p ${builddir}

cd ${builddir}
wget --output-document=scipy-${VERSION_SCIPY}.tar.gz \
        http://sourceforge.net/projects/scipy/files/scipy/${VERSION_SCIPY}/scipy-${VERSION_SCIPY}.tar.gz/download
tar -xvvzkf scipy-${VERSION_SCIPY}.tar.gz
cd scipy-${VERSION_SCIPY}
python setup.py build
python setup.py install --home=${installdir}

matplotlib

• First you need to install scipy as shown above and make sure PYTHONPATH points to the new numpy installation.

VERSION_MATPLOTLIB=1.0.1
installdir="${HOME}/opt"
builddir="/scratch/${USER}/build/matplotlib"

export PYTHONPATH=${installdir}/lib/python

mkdir -p ${builddir}

cd ${builddir}
wget --output-document=matplotlib-${VERSION_MATPLOTLIB}.tar.gz \
        http://sourceforge.net/projects/matplotlib/files/matplotlib/matplotlib-${VERSION_MATPLOTLIB}/matplotlib-${VERSION_MATPLOTLIB}.tar.gz/download
tar -xvvzkf matplotlib-${VERSION_MATPLOTLIB}.tar.gz
cd matplotlib-${VERSION_MATPLOTLIB}
python setup.py build
python setup.py install --home=${installdir}

nose

• This module is required to run e.g. the numpy and scipy test suites.

VERSION_NOSE=1.0.0
installdir="${HOME}/opt"
builddir="/scratch/${USER}/build/nose"

export PYTHONPATH=${installdir}/lib/python

mkdir -p ${builddir}

cd ${builddir}
wget http://somethingaboutorange.com/mrl/projects/nose/nose-${VERSION_NOSE}.tar.gz
tar -xvvzkf nose-${VERSION_NOSE}.tar.gz
cd nose-${VERSION_NOSE}
python setup.py build
python setup.py install --home=${installdir}

CategoryLXSW