Travis-ci matplotlib dependency and python3

Question

I am trying to setup a travis continuous build system with my project, which has numpy, scipy and matplotlib in its dependencies. I am targeting python 3.3.

In my .travis.yml script I am installing numpy and scipy from apt-get, as well as (to be sure) from pip (only numpy). Unfortunatelly, matplotlib build still says that numpy is missing from deps. I tried almost all the methods found on the WEB but most of them do not work (they are outdated I think).

language: python                                                                                                                                                                                                                    
python:                                                                                                                                                                                                                             
  - "3.3"                                                                                                                                                                                                                           
install:                                                                                                                                                                                                                            
  - pip install numpy                                                                                                                                                                                                               
  - pip install colorama
  - pip install matplotlib
  - pip install nose                                                                                                                                                                                                                
script: nosetests                                                                                                                                                                                                                   
virtualenv:                                                                                                                                                                                                                         
  system_site_packages: true                                                                                                                                                                                                        
before_install:                                                                                                                                                                                                                     
  - sudo apt-get update -qq                                                                                                                                                                                                         
  - sudo apt-get install -qq python3-numpy python3-scipy  

Below is the interesting part of travis log. It says that dependence is not satisfied, yet pip command can see numpy installed already from apt.

BUILDING MATPLOTLIB
            matplotlib: 1.2.0
                python: 3.3.2 (default, May 16 2013, 18:32:41)  [GCC 4.6.3]
              platform: linux

REQUIRED DEPENDENCIES
                 numpy: no
                        * You must install numpy 1.4 or later to build
                        * matplotlib.
Complete output from command python setup.py egg_info:
basedirlist is: ['/usr/local', '/usr']                                                                                                                                                              

Answer 1

If you don't need to test against multiple python versions, the easiest trick is to tell travis that your language is c and then install everything from apt-get. This gets around all of the issues with system_site_packages and virtualenv.

This library, for instance, uses travis-ci for testing and depends on the full scipy stack (numpy, scipy, matplotlib, pytables, pandas, etc), which is installed via apt with language=c.

https://github.com/rmcgibbo/mdtraj/blob/master/.travis.yml

Answer 2

Apt-get, Robert McGibbon's suggestion, is still apparently quite slow.

Here's an approach from Dan Balchard using Miniconda, that will pre-install matplotlib and the rest of the scipy stack on your Travis CI test machine. Here's the full .travis.yml file:

language: python
python:
  - 2.7
  - 3.3
notifications:
  email: false

# Setup anaconda
before_install:
  - wget http://repo.continuum.io/miniconda/Miniconda-latest-Linux-x86_64.sh -O miniconda.sh
  - chmod +x miniconda.sh
  - ./miniconda.sh -b
  - export PATH=/home/travis/miniconda/bin:$PATH
  - conda update --yes conda
  # The next couple lines fix a crash with multiprocessing on Travis and are not specific to using Miniconda
  - sudo rm -rf /dev/shm
  - sudo ln -s /run/shm /dev/shm
# Install packages
install:
  - conda install --yes python=$TRAVIS_PYTHON_VERSION atlas numpy scipy matplotlib nose dateutil pandas statsmodels
  # Coverage packages are on my binstar channel
  - conda install --yes -c dan_blanchard python-coveralls nose-cov
  - python setup.py install

# Run test
script:
  - nosetests --with-cov --cov YOUR_PACKAGE_NAME_HERE --cov-config .coveragerc --logging-level=INFO

# Calculate coverage
after_success:
  - coveralls --config_file .coveragerc