Class 7 - testing

Python package at the end of class: testing.zip.

All scientific code is tested; untested code is not scientific. We covered automated testing and how it can make the task of code testing better/easier.

Topics covered:

1. General discussion about testing software and scientific software in particular: manual, exploratory vs automated testing, end-to-end tests, unit tests.
2. Testing simple functions.
3. Testing with pytest: fixtures, marks (skipping slow tests), exceptions.
4. Supplementing pytest with hypothesis, including numpy extension.
5. Briefly about Numpy support for testing in numpy.testing.
6. Briefly about using tox for testing different Python interpreter and library (e.g. numpy, scipy) versions. (To maintain different python versions use pyenv or conda under Mac OS and Linux; use conda under Windows.) versions