Travis CI is a Continuous Integration service that will run tests for your project anytime someone makes a change or opens a pull request. However, every time a build starts on Travis, all the dependencies are installed from scratch in a clean environment. Travis does include Python, but if you try and install for example Numpy, Scipy, and Matplotlib from source, the installation could take a while (over 20 minutes) for each build.
This repository contains a simple .travis.yml file to get you started with
testing your Python project on Travis CI using
conda to speed up the installation of dependencies.
Setting up a build with for example Python, Numpy, Scipy, Matplotlib, and
Astropy takes only a minute! This is possible because
conda installs pre-compiled binary packages.
To get started, copy over the .travis.yml file here to your repository, and
add/remove dependencies or builds in the matrix if needed. Then go to the
Travis CI website, sign in with your GitHub account,
then go to Accounts and turn on Travis for the repository you want to test.
Next time you push a change or open a pull request on your repository, Travis
will run the tests!
Note that in the example .travis.yml file, the command to run the tests is
currently set to py.test (in the script section) but you can change
this to anything you like.
You can find out more information about using conda on Travis in
this official conda blog post blog
post - their example .travis.yml file is a little different, but both will
do the job. The one in this repository is a little simpler overall and includes
more inline comments.
