Saturday, June 8, 2013

Setting up Python/Pylab environment on machine with no root access

Often you will get the opportunity to run your code on a cluster, or even simply a guest computer, where you don't have root access (sudo does not work). The great thing about Python is that Python and any modules can be completely installed in user space (in your account) without disturbing any one else. In the computer I am accessing (Partners' HPC setup) they are hip and have all versions of Python. In case your setup doesn't you first want to do:
cd /tmp #(We have write access here)
curl -O #Or whatever the version you want
tar -xzf Python-2.7.5.tar.bz2 
cd Python-2.7.5
./configure --prefix=$HOME/local/Python #Or where ever you want it
make install
Make sure this directory is in your path such that you can call python.
#Add this path to .bash_profile
Python has a very neat, principled, way of storing its infrastructure (modules) that is detailed here. A quick way to figure out where this is is to do
python -m site --user-site
We need to add this to our path too. In my case is was ~/.local/bin
#Add this path to .bash_profile
Next you will want pip, which is a more modern package manager than easy_install, but first you'll want setup tools:
wget -O - | python

cd /tmp #(We have write access here)
curl -O
tar -xzf pip-1.3.tar.gz
cd pip-1.3/
python install --user
Python informs us pip is now installed under
. The magic from now on is to simply add --user to our pip install commands to install packages in our userspace. Let's try this out:
pip install matplotlib --user
pip install ipython --user
pip install --upgrade nose --user #(The server had an older version of nose)
And we have our environment. The HPC cluster I went into already had numpy and scipy, but you may need to install those as well.

No comments:

Post a Comment