Skip to content

TUTORIAL.md

Latest commit

 

History

History
57 lines (48 loc) · 3.73 KB

TUTORIAL.md

File metadata and controls

57 lines (48 loc) · 3.73 KB

Continuous integration tutorial

This tutorial presents how to use continuous integration (CI) in a Python code development, using the pytest and Travis. Through a few simple exercises, you'll go through different features of CI.

pytest module of Python

pytest is a framework that makes building simple and scalable tests easy.

You can install it using pip (pip install -U pytest) or in a conda environment (conda install pytest). You can run pytest within a terminal (linux and osx) or in the anaconda prompt (Windows).

Exercise 0: Hello World!

In this exercise, you will write a first test in a file test_hello_world.py.

  • In a file test_hello_world.py, write a python function that prints "Hello World!" and returns 0.
  • In that same file, write a new function test_hello_world() that call your hello world function and asserts that it returned 0.
  • Test your file running pytest test_hello_world.py.
  • Questions:
    • What happens if you run pytest . ?
    • And if you rename the file test_hello_world.py to hello_world.py ?
    • And if you rename the functions ?

Exercise 1: The Fibonacci sequence

The Fibonacci sequence is a sequence of natural numbers where each number is the sum of the two previous numbers in the sequence, starting by 0 and 1:

F[0] = 0
F[1] = 1
F[n] = F[n-1] + F[n-2]
  • In a file fibonacci.py, write a function that returns the n th Fibonacci number.
  • In a file test_fibonacci.py, write a function test_fibonacci() that asserts the 5th Fibonacci number is a positive natural number.
  • Modify test_fibonacci() to test_fibonacci(n) that checks the n th Fibonacci number, and test it for different values of n.
    • You'll need to use @pytest.mark.parametrize('n', [0, 1, 5, -1]) as a decorator to your function.
    • What happens for a negative parameter n ?

Exercise 2: The Fibonacci sequence revisited

  • You now want to write a function strict_fibonacci(n) that raises an error if parameter n is not a natural number.
  • You can also test this function, checking that it raises an error as expected for negative n:
    • You'll need to use such decorator: @pytest.mark.parametrize('n', [1, pytest.param(-1, marks=pytest.mark.xfail(strict=True))])

Using github and Travis

Using Travis CI, you can automatically test every commit you push on github.

How to get started with Travis (see more info in the Travis tutorial)?

  • You can find Travic CI (and other Github Apps) on the Github Marketplace.
  • Install and configure Travis, selecting the repository you want to test.
  • Add a .travis.yml file to your repository to tell Travis CI what to do.
    • Travis can test your code using different versions of Python. See this example of .travis.yml
    • Testing Python under OSX is currently broken. You can't do it as for the linux example above, but you can still test an OSX Python by installing it with conda. To use any conda environment, you can also use such approach, with this .travis.yml
  • Now, each time you'll push a commit on github, your full code will be tested.
  • You can setup Travis in your settings to only test master branch, every branch, the pull requests, etc.

The AppVeyor CI tool

To test your code on Windows, you can use AppVeyor, which is similar to Travis.