Tjelvar Olsson     About     Posts     Feed     Newsletter

I'm currently working on a computing book aimed at biologists. Check out the Biologist's Guide to Computing!

How to continuously test your Python code on Windows using AppVeyor

In the previous post I illustrated how to setup continuous integration testing of your Python code using Travis CI. Travis CI is great when working on Linux. However, what can you do if you wanted to setup automated continuous integration testing on Windows?

To me, a Linux enthusiast, this problem sounded almost insurmountable…

AppVeyor to the rescue

However, it turns out that AppVeyor has provided a service for solving this problem.

One simply needs to create an appveyor.yml file to configure the running of the test suite. The code below creates a testing matrix for running the test suite on 32-bit Python 2.7, 3.3 and 3.4 using the nosetests test runner.

build: false

    - PYTHON: "C:\\Python27"
      PYTHON_VERSION: "2.7.8"
      PYTHON_ARCH: "32"

    - PYTHON: "C:\\Python33"
      PYTHON_VERSION: "3.3.5"
      PYTHON_ARCH: "32"

    - PYTHON: "C:\\Python34"
      PYTHON_VERSION: "3.4.1"
      PYTHON_ARCH: "32"


  - "%PYTHON%/Scripts/pip.exe install nose"
  - "%PYTHON%/Scripts/pip.exe install coverage"

  - "%PYTHON%/Scripts/nosetests"

Note that we use pip to install the nose and coverage packages before we run the test suite.

Commit and push this file and login to AppVeyor using your GitHub account. Sync your GitHub repositories and then select the project you want AppVeyor to run continuous integration testing on.

Job done!

Using Minconda to test projects that depend on the numpy/scipy stack

Again testing projects that depend on numpy and scipy present problems in that these packages take too long to build from scratch. However, just like in the previous post we can make use of Miniconda.

In fact the kind people at AppVeyor have already deployed Minicoda to their build workers (

So to test a project that depends on numpy and scipy one can simply use the appveyor.yml file below.

build: false

      MINICONDA: C:\Miniconda
      MINICONDA: C:\Miniconda3


  - "set PATH=%MINICONDA%;%MINICONDA%\\Scripts;%PATH%"
  - conda config --set always_yes yes --set changeps1 no
  - conda update -q conda
  - conda info -a
  - "conda create -q -n test-environment python=%PYTHON_VERSION% numpy scipy nose"
  - activate test-environment
  - pip install coverage

  - nosetests

The script above installs numpy, scipy and nose using the Conda package manager. However, the Conda package manager does not contain the coverage package. We therefore install that using pip instead after the virtual environment has been activated.

The fact that Miniconda is included in the AppVeyor makes it trivial to test Python code with scientific dependencies.

Great stuff!

See also