A terse boilerplate for Python packages. Where other boilerplates contain a great many features such as a CLI, docs, code coverage and varying opinions on what a unit test should look like, this one aims to be lean and mean. I use this on my own projects.
The package structure is:
├── packagename
│ ├── __init__.py
│ ├── __main__.py
│ └── somemodule.py
└── tests
│ ├── __init__.py
│ ├── test_myclass.py
│ └── test_mysecondclass.py
├── setup.py
├── .travis.yml
├── .gitignore
├── README.md
├── LICENSE
└── VERSION
Package requirements are handled using setup.py. They will be automatically installed on install or during unit testing
Testing is set up using the built in Unit Test framework in Python
Run the tests with python setup.py test in the root directory or use the handy test.sh script in the /scripts folder.
There is a .travis.yml file that will execute the build and the unit tests against several versions of Python.
git clone https://github.com/dagostinelli/python-package-boilerplate
cd python-package-boilerplate
virtualenv venv
source venv/bin/activate
cd scripts
./build.sh
./test.sh
Then begin renaming things and keep running those unit tests as you go.
This example shows how to do a command line interface (CLI). You can try it like this:
git clone https://github.com/dagostinelli/python-package-boilerplate
cd python-package-boilerplate
python -m packagename
or like this
git clone https://github.com/dagostinelli/python-package-boilerplate
cd python-package-boilerplate
pipsi install .
packagename
I enjoy filesystem monitors that look for files as I change them and then automatically execute the unit tests in the background. I haven't found a cross-platform way to do that yet, so it's not in this boilerplate. For interested parties running Linux, you can do it in a make file. Create a make watch target in your Makefileand call inotifywatch there.