Install Python packages on Ubuntu 14.04


This post is more than 5 years old. While math doesn't age, code and operating systems do. Please use the code/ideas with caution and expect some issues due to the age of the content. I am keeping these posts up for archival purposes because I still find them useful for reference, even when they are out of date!


There is an updated post for Ubuntu 16.04-- python setup on Ubuntu 16.04 , check it out for current information.

In this post I will document my setup of Python 2.7.6 in Ubuntu 14.04. Of course, the base Python is installed by default, but both Python 2.7.6 and Python 3.4 are available. Try the following in the terminal:

$ python --version
Python 2.7.6
$ python2 --version
Python 2.7.6
$ python3 --version
Python 3.4.0

As you can see, using python points to Python 2.7.6 by default. However, python2 and python3 can be used to access the desired version. I will focus on installing packages for Python 2.7.6 here.


In the past I have installed Python packages:

  1. Using the Ubuntu repository:
    $ sudo apt-get install packagename
  2. Or, from a git/svn repository:
    $ sudo python install

Approach 1 has many advantages for Python users who don't need to have the latest versions of every package. In particular, all of the package dependencies including other Python packages, linear algebra libraries, etc. are also installed automatically. As a result, if you are new to Ubuntu and Python, strategy 1 is the way to go.

I will take a different tact, using pip to install, upgrade, and remove packages. Also, I will install all Python packages as a user, that is, no use of sudo. This makes it easy to use the same install procedure on a machine where I don't have sudo privileges--say an Ubuntu cluster. However, I will need sudo to install non-Python libraries, Fortran compilers, etc. that the Python packages employ. On a cluster, the SysAdmin would have to to do this part for me and other users.


June 1st, 2015: Recently use of pip on Ubuntu 14.04 has started to issue a warning that ends with InsecurePlatformWarning. After some searching around, I've found that this is related to SSL and the urllib3 in Python 2.7.6, the version on Ubuntu 14.04-- see here if you want the details. As suggested in the discussion linked above, this can be fixed with the following installs (I'll use the --user switch, as in the examples below)

$ pip install --user pyopenssl ndg-httpsclient pyasn1

With that we're secured and the warnings go away. If you are just starting out, try installing pip, as below, and return to this install if use of pip gives you warnings.


Of course, the starting point is to get pip installed. Official instructions are also available for installing pip . pip depends on setuptools, but we can install both using the script, as described at the install link. To be concrete, I did the following:

$ cd ~/Downloads
$ curl -O
$ python --user

If you don't have curl installed, this can be remedied using:

$ sudo apt-get install curl

Because we have chosen local installation, the path ~/.local/bin has to be added to our path. To do that, add the following to the end of your ~/.bashrc file::

# include .local/bin for local python scripts
export PATH=~/.local/bin:$PATH

Then, source ~/.bashrc:

$ source ~/.bashrc

Try the following to see if you get similar results and to make sure the basic setup is working:

$ which pip
$ pip --version
pip 1.5.6 from /home/cstrelioff/.local/lib/python2.7/site-packages (python 2.7)

Of course, your username should be in the path, but the output should look something like the above.


Another major tool for Python 2.7 project management is virtualenv . This package allows the user to create many virtual Python environments, with different packages installed, and to activate and deactive these environments whenever the user desires. This is extremely useful for developers who want to create a minimal environment for their application.

The virtualenv installation is simple with pip (again, I'm doing a user install with no sudo):

$ pip install --user virtualenv

To test it out, see if you get something like the following:

$ virtualenv --version
$ pip show virtualenv
Name: virtualenv
Version: 1.11.6
Location: /home/cstrelioff/.local/lib/python2.7/site-packages

Now that virtualenv is installed, there will be two paths forward for the rest of the Python installs:

  1. Keep installing as a user-- I'll use this approach for the reasons discussed above.
  2. If you have admin permissions you can install all packages globally using a command like:
    $ sudo pip install packagename
  3. Create a virtual environment and install everything there to have a completely isolated Python environment-- see virtualenv and virtualenvwrapper on Ubuntu 14.04 for an example of how to take this approach.

Ubuntu dependencies

A variety of Ubuntu-specific packages are needed by Python packages. These are libraries, compilers, fonts, etc. I'll detail these here along with install commands. Depending on what you want to install you might not need all of these.

  • General development/build:
    $ sudo apt-get install build-essential python-dev
  • Compilers/code integration:
    $ sudo apt-get install gfortran
    $ sudo apt-get install swig
  • Numerical/algebra packages:
    $ sudo apt-get install libatlas-dev
    $ sudo apt-get install liblapack-dev
  • Fonts (for matplotlib)
    $ sudo apt-get install libfreetype6 libfreetype6-dev
  • More fonts (for matplotlib on Ubuntu Server 14.04-- see comment at end of post) -- added 2015/03/06
    $ sudo apt-get install libxft-dev
  • Graphviz for pygraphviz, networkx, etc.
    $ sudo apt-get install graphviz libgraphviz-dev
  • IPython require pandoc for document conversions, printing, etc.
    $ sudo apt-get install pandoc
  • Tinkerer dependencies
    $ sudo apt-get install libxml2-dev libxslt-dev zlib1g-dev
That's it, now we start installing the Python packages.


numpy is one of the fundamental numerical packages in Python. To install using pip type:

$ pip install --user numpy

This will result in a fair amount of compiling followed by a note that the package was successfully installed. If not, make a note of the error. Often this results from not having libraries and/or compilers installed (see above).

Information about the installation location and the version can be obtained with the following:

$ pip show numpy
Name: numpy
Version: 1.8.1
Location: /home/cstrelioff/.local/lib/python2.7/site-packages

You should also be able to start python at the terminal and import numpy without complaint:

>>> import numpy as np
>>> print np.__version__
>>> exit()


scipy has many useful mathematical utilities, complementing numpy . Installation is accomplished with:

$ pip install --user scipy

Again, expect lots of compiling! As with numpy , try:

$ pip show scipy
Name: scipy
Version: 0.14.0
Location: /home/cstrelioff/.local/lib/python2.7/site-packages

and, loading python:

>>> import scipy
>>> print scipy.__version__
>>> exit()


matplotlib is one of the main plotting packages for Python and many other packages use the utilities. Install with:

$ pip install --user matplotlib

If you look carefully, the completion of the installation will say:

Successfully installed matplotlib python-dateutil tornado pyparsing nose backports.ssl-match-hostname Cleaning up...

So, matplotlib installs a variety of Python-dependencies. As usual, try:

$ pip show matplotlib
Name: matplotlib
Version: 1.3.1
Location: /home/cstrelioff/.local/lib/python2.7/site-packages
Requires: numpy, python-dateutil, tornado, pyparsing, nose

Finally try a simple plot:

>>> import matplotlib.pyplot as plt
>>> plt.plot([1,2,3,4])
[<matplotlib.lines.Line2D object at 0x7f13a8571890>]
>>> plt.ylabel('some numbers')
<matplotlib.text.Text object at 0x7f13a85c47d0>
>>> exit()

A plot should open in a new window when is executed.


sympy is a computer algebra system for Python. Install with pip using:

$ pip install --user sympy

Again, installation information from pip is obtained with:

$ pip show sympy
Name: sympy
Version: 0.7.5
Location: /home/cstrelioff/.local/lib/python2.7/site-packages

Finally, following the sympy tutorial , start Python and try:

>>> from sympy import symbols
>>> x, y = symbols('x y')
>>> expr = x + 2*y
>>> expr
x + 2*y
>>> expr + 1
x + 2*y + 1
>>> expr - x
>>> exit()



Next, we install IPython (including notebooks), which has become a major tool for sharing python projects in an interactive format. To install we use:

$ pip install --user ipython[notebook]

At the end, we get the message:

Successfully installed ipython jinja2 pyzmq markupsafe Cleaning up...

showing that jinja2, pyzmq and markupsafe have also been installed. Get install information from pip:

$ pip show ipython
Name: ipython
Version: 2.1.0
Location: /home/cstrelioff/.local/lib/python2.7/site-packages

Now, try:

$ ipython

which launches the IPython terminal. Notice the IPython version is provided and the prompt looks different from the normal >>> Python prompt (see the IPython documentation for more information):

Python 2.7.6 (default, Mar 22 2014, 22:59:56) 
Type "copyright", "credits" or "license" for more information.

IPython 2.1.0 -- An enhanced Interactive Python.
? -> Introduction and overview of IPython's features.
%quickref -> Quick reference.
help -> Python's own help system.
object? -> Details about 'object', use 'object??' for extra details.

In [1]: import numpy as np

In [2]: print np.__version__

In [3]: exit()

Finally, IPython_ notebook can be launched with the command:

$ ipython notebook

This launches a web browser and you should see the IPython_ notebook interface. You can create a new notebook and work away. To shutdown the server, back at the terminal where you launched the notebook, type cntrl-c and then y when prompted:

Shutdown this notebook server (y/[n])? y
2014-06-04 16:29:04.033 [NotebookApp] CRITICAL | Shutdown confirmed
2014-06-04 16:29:04.033 [NotebookApp] Shutting down kernels

That's it, you're now an IPython notebook user!


pygraphviz is a Python interface to the graphviz visualization code that can be used by itself but is also employed by networkx and other packages. Be sure that graphviz and its developer libraries are installed (see Ubuntu Dependencies above) and install pygraphviz using:

$ pip install --user pygraphviz

Get install information from pip :

$ pip show pygraphviz
Name: pygraphviz
Version: 1.2
Location: /home/cstrelioff/.local/lib/python2.7/site-packages

Also, try:

>>> import pygraphviz
>>> print pygraphviz.__version__
>>> exit()


networkx is a Python package for building, analyzing, and visualizing graphs/networks. There are a variety of dependencies, all of which we have installed above. So, install with:

$ pip install --user networkx

Get install information from pip :

$ pip show networkx
Name: networkx
Version: 1.8.1
Location: /home/cstrelioff/.local/lib/python2.7/site-packages

Try a simple example:

>>> import networkx as nx
>>> G = nx.Graph()
>>> G.add_edge(1,2)
>>> G.add_edge(2,3)
>>> import matplotlib.pyplot as plt
>>> nx.draw(G)
>>> exit()

With matplotlib and pygraphviz installed (see above), this code should create a very simple graph and show it in a new window when is executed.


pandas is a Python packaged focused on data -- reading, writing, manipulating, etc. There are a variety of pandas dependencies : required, recommended and/or optional. We'll focus on the first two categories.

The required dependencies are numpy (installed above), python-dateutil (installed above with matplotlib), and pytz (we will let pip install with pandas). However, let's install the recommended dependencies:


Install numexpr with:

$ pip install --user numexpr

After install we get:

$ pip show numexpr
Name: numexpr
Version: 2.4
Location: /home/cstrelioff/.local/lib/python2.7/site-packages
Requires: numpy


Install bottleneck using:

$ pip install --user Bottleneck

After install we get:

$ pip show Bottleneck
Name: Bottleneck
Version: 0.8.0
Location: /home/cstrelioff/.local/lib/python2.7/site-packages

We can also import both packages in Python and print the package version to make sure that basic usage seems okay:

$ python -c "import numexpr;print numexpr.__version__"
$ python -c "import bottleneck;print bottleneck.__version__"

Finally, for pandas , we install the main package:

$ pip install --user pandas

After some downloading and compiling we get (showing that both pandas *and* pytz were installed, as expected):

Successfully installed pandas pytz Cleaning up...

Use pip_ to check the installation information:

$ pip show pandas
Name: pandas
Version: 0.14.0
Location: /home/cstrelioff/.local/lib/python2.7/site-packages
Requires: python-dateutil, pytz, numpy


If you import pandas , an error about openpyxl (a package for working with Excel 2007 files) will be issued:

>>> import pandas
/home/cstrelioff/.local/lib/python2.7/site-packages/pandas/io/ UserWarning: Installed openpyxl is not supported at this time. Use >=1.6.1 and <2.0.0.
.format(openpyxl_compat.start_ver, openpyxl_compat.stop_ver))
>>> exit()

The error says that openpyxl needs to be at least version 1.6.1 and less than 2.0.0. Strange, this package is listed as optional. Oh well, let's install an appropriate version. If we just use pip to install the current version it will be too high. So, I installed as follows:

openpyxl 1.8.6

$ pip install --user openpyxl==1.8.6

This install forces the use an appropriate version. Now, try importing pandas and we get:

>>> import pandas
>>> print pandas.__version__
>>> import openpyxl
>>> print openpyxl.__version__
>>> exit()

Yay(!) we can import pandas_ (and openpyxl) without complaints.

Finally, before leaving pandas , I will mention that there are a variety of optional pandas dependencies that you might want to consider as well. I won't consider them in this post.


pymc is a really nice MCMC package for Python. I have used it on several projects with great success. Installation with pip follows the usual format:

$ pip install --user pymc

Get install information:

$ pip show pymc
Name: pymc
Version: 2.3.2
Location: /home/cstrelioff/.local/lib/python2.7/site-packages

Starting Python you should also be able to get:

>>> import pymc
>>> print pymc.__version__
>>> exit()


statsmodels provides some nice statistics methods. Before installing statsmodels itself, we must install dependencies, which will likely be usesul in any case: patsy and cython .


patsy is a package for describing statistical models in R-like format. Install with:

$ pip install --user patsy

We can see where pip installed patsy :

$ pip show patsy
Name: patsy
Version: 0.2.1
Location: /home/cstrelioff/.local/lib/python2.7/site-packages
Requires: numpy

and try importing patsy in a Python session:

>>> import patsy
>>> print patsy.__version__
>>> exit()


cython allows for wrapping of c++ code. Install with:

$ pip install --user Cython

Check with pip:

$ pip show Cython
Name: Cython
Version: 0.20.1
Location: /home/cstrelioff/.local/lib/python2.7/site-packages

and importing in a Python session:

>>> import cython
>>> print cython.__version__
>>> exit()

Finally, install statsmodels with pip:

$ pip install --user statsmodels

Show install info with pip:

$ pip show statsmodels
Name: statsmodels
Version: 0.5.0
Location: /home/cstrelioff/.local/lib/python2.7/site-packages

and try an import:

>>> import statsmodels
>>> print statsmodels.__version__
>>> exit()

Okay, that's patsy , cython and statsmodels .


CMPy is a package for Computational Mechanics in Python developed in the Crutchfield Lab at UC Davis. Currently the package is developed, using git for version control, but is not publicly available. However, I will document the install here because:

  1. It's useful for people at UCD (or collaborating with people at UCD)
  2. This is an example of installation of a Python package in a folder on the local machine

I start by showing that I have cloned the CMPy package to the ~/gitlocal/cmpy/ directory. You can see the file when I list the directory contents:

$ ls ~/gitlocal/cmpy/
apps build CHANGES.txt cmpy data docs gallery LICENSE.txt old_doc pylintrc README.txt scripts src

We do the install with pip, using the -e switch to show the location of the package code:

$ pip install --user -e ~/gitlocal/cmpy/
Obtaining file:///home/cstrelioff/gitlocal/cmpy
Running (path:/home/cstrelioff/gitlocal/cmpy/ egg_info for package from file:///home/cstrelioff/gitlocal/cmpy

Installing collected packages: CMPy
Running develop for CMPy

Creating /home/cstrelioff/.local/lib/python2.7/site-packages/CMPy.egg-link (link to .)
Adding CMPy 1.0dev to easy-install.pth file

Installed /home/cstrelioff/gitlocal/cmpy
Successfully installed CMPy
Cleaning up...

Note that the path to the CMPy_ directory is added to easy-install.pth, a file that Python consults to find CMPy. Finally, we show the pip information:

$ pip show cmpy
Name: CMPy
Version: 1.0dev
Location: /home/cstrelioff/gitlocal/cmpy

Again, note that the location is ~/gitlocal/cmpy/, instead of ~/.local/lib/python2.7/site-packages/, due to the -e tag. This is why the addition to the easy_install.pth file (above) was needed.


Aug 21st, 2014 : A note on updating this local installation is in order. Recently a change in code was made that affected underlying c code that is incorporated using cython. I pulled the repository changes using:

$ cd ~/gitlocal/cmpy/
$ git pull

To try and update the install I did:

$ pip install --user -e ~/gitlocal/cmpy/

This ran the file but did not recompile the modified c code. To get this to work I had to remove the build directory, build in place and install again:

$ cd ~/gitlocal/cmpy/
$ rm -r build/
$ python build_ext -i --cython
$ pip install --user -e ~/gitlocal/cmpy/

Is there a better way to do this? Let me know in the comments below.


restview is a Python package that processes reStructuredText and launches a web browser for viewing. Each time the browser is refreshed, the underlying rst document will be re-processed and displayed-- very nice for working on Python docmentation or any rst document. Installation goes as usual:

$ pip install --user restview

We can see what was installed:

$ pip show restview
Name: restview
Version: 2.0.5
Location: /home/cstrelioff/.local/lib/python2.7/site-packages
Requires: docutils, pygments

As you can see from above, docutils and pygments will be installed if they are not already installed.

To process an rst document named test.rst type:

$ restview test.rst

Check restview for more examples.


tinkerer is a blogging environment for Pythonistas that is built on Sphinx , a Python documentation tool. Blog entries are written in reStructuredText and rendered as static html. Of course, this is also the tool I use for this blog. Before moving to our usual pip install, we have to take care of some Ubuntu 14.04 Python dependencies (see above). Assuming these requirements are available, tinkerer is installed with the usual:

$ pip install --user tinkerer

We can check the install information with:

$ pip show tinkerer
Name: Tinkerer
Version: 1.4.2
Location: /home/cstrelioff/.local/lib/python2.7/site-packages
Requires: Jinja2, Sphinx, Babel, pyquery

Note that requirements Jinja2 , Sphinx , Babel and pyquery are also installed automatically. A quick start to getting a blog up and running (at least the generation of posts, pages and generating the html output) is available here.


Pweave is a tool for literate programming with Python. This tool allows me to write blog posts about Python using a .Pnw file that contains reStructuredText , along with special Pweave commands, and have the Python code evaluated and output included in the .rst output file-- see the example here. This is a really nice tool to avoid typos in code and to make sure that what you're talking about actually works! I should note that IPython notebooks can also do this by exporting to reStructuredText . In any case, I will trying out both of these tools for future posts.

The install of Pweave goes as usual:

$ pip install --user Pweave

Check the install with:

$ pip show Pweave
Name: Pweave
Version: 0.21.2
Location: /home/cstrelioff/.local/lib/python2.7/site-packages


scikit-learn is the probably the most well-known and feature-complete package for machine learning tasks in Python. There are a number of dependencies that need to be installed ( numpy , scipy , python-dev, etc see scikit-learn installation for more information) that have already been installed above. So, we install using pip, as usual:

$ pip install --user scikit-learn

Then we can check the installed version and location using:

$ pip show scikit-learn
Name: scikit-learn
Version: 0.15.1
Location: /home/cstrelioff/.local/lib/python2.7/site-packages

That's it, machine-learn away!