Python Development with Virtual Environments

From BC Wiki
Jump to: navigation, search

Topics to cover:

  • Pip
  • Virtualenv
  • Virtualenvwrapper

Python Install Python (PIP)[edit]

You can use pip freeze or pip list to generate a list of the Python packages installed in your current environment. You can output this list to a file and use that file to install the exact same packages in another environment.

Virtualenvwrapper[edit]

https://virtualenvwrapper.readthedocs.org/en/latest/

See existing environments—and switch between them—with workon (has tab completion).

Make new environments with mkvirtualenv.

Once in a virtual environment, use this to update all Python packages:

pip freeze --local | grep -v '^\-e' | cut -d = -f 1 | xargs -n1 pip install -U

Development Environment Examples[edit]

All of these examples start from a completely fresh copy of the listed operating system. I create and test these configurations, and do most of my development and troubleshooting work, by cloning a virtual machine running fully-up-to-date Ubuntu 16.04 with the user interface configured exactly how I like it; the cloned VM uses a linked disk tied to the parent VM that only records changes, so I can create a new clean VM to experiment with in just a few seconds and it doesn't take up very much hard drive space.

Ubuntu 16.04 - Scientific Computing[edit]

Prepare the virtual environment tools:

sudo apt install python-pip
pip install --upgrade pip
pip install virtualenvwrapper
nano ~/.bashrc
# add this to the end of the file:
  # setup for virtualenvwrapper
  export WORKON_HOME=$HOME/.virtualenvs
  export PATH=$PATH:$HOME/.local/bin
  source /home/brandon/.local/bin/virtualenvwrapper.sh
source ~/.bashrc 

set up a basic virtual environment:

mkvirtualenv py2sci
pip install requests[security]
pip install numpy scipy jupyter

Now we have to install some system-wide stuff so that we can use the various plotting backends in matplotlib. We have to use `apt install` instead of `pip install` because these things are not Python packages.

Installing wxpython_phoenix should get wx-based backends working:

pip install --upgrade --trusted-host wxpython.org --pre -f http://wxpython.org/Phoenix/snapshot-builds/ wxPython_Phoenix

Installing pyside should get qt4-based backends working:

sudo apt-get install build-essential cmake libqt4-dev libxml2-dev libxslt1-dev python-dev qtmobility-dev python-pip

pip install pyside

Installing cairocffi should get cairo-based backends working:

pip install cairocffi

Installing tkinter and the development headers should get tk-based backends working:

sudo apt-get install python-tk tk-dev

Now you can install matplotlib, either using pip:

pip install matplotlib

or from source pulled from Github:

sudo apt-get install dvipng
git clone https://github.com/matplotlib/matplotlib.git
cd matplotlib
python setup.py install

If you'd like the gtk3-based backends to work as well, you'll need to install some stuff system-wide; after you've copied it into your virtual environment folder, you can get uninstall it:

sudo apt install python-gobject python-gi-cairo
cp -R /usr/lib/python2.7/dist-packages/gi ~/.virtualenvs/mplpip/lib/python2.7/site-packages/
cp -R /usr/lib/python2.7/dist-packages/glib ~/.virtualenvs/mplpip/lib/python2.7/site-packages/
cp -R /usr/lib/python2.7/dist-packages/gobject ~/.virtualenvs/mplpip/lib/python2.7/site-packages/
cp -R /usr/lib/python2.7/dist-packages/gtk-2.0 ~/.virtualenvs/mplpip/lib/python2.7/site-packages/
cp -R /usr/lib/python2.7/dist-packages/cairo ~/.virtualenvs/mplpip/lib/python2.7/site-packages/
sudo apt purge python-cairo python-gi python-gi-cairo python-gobject python-gobject-2

Launch Jupyter Notebook with `jupyter notebook`. Create a new notebook and check out the available matplotlib plotting backends with %matplotlib -l<?code>; select the one you'd like to use with <code>%matplotlib <backend> before importing matplotlib's pyplot with from matplotlib import pyplot as plt.

matplotlib is very well-documented; check out what's new in the latest release, read more about how to plot and choose a plotting backend, and see examples of figures in the matplotlib gallery.

Caltech has a great intro to Jupyter notebooks, and you can even download their example notebook and play with it yourself.