Starting a new Python project from the scratch is boring and error prone: * First create a setup.py script * Create documentation * Provide installation instructions * A README file * Write tests * etc, etc, etc.

This time-consuming and error prone work gives little satisfaction, but is necessary to make your project a good citizen in the open source community.

python-boilerplate produces skeletons for your Python projects so you can get up and running fast. It is influenced by this blog post: http://jeffknupp.com/blog/2013/08/16/open-sourcing-a-python-project-the-right-way/, although we do not follow these recommendations by the letter.

We start a python-boilerplate skeleton by calling the python-boilerplate init command on the root directory of your project. It creates the following tree:

.
|- .coveragerc
|- .gitignore
|- .travis.yml
|- LICENSE
|- MANIFEST.in
|- INSTALL.rst
|- README.rst
|- VERSION
|- requirements.txt
|- setup.py
|- tasks.py
|- tox.ini
|- docs/
|   |- index.rst
|   |- apidoc.rst
|   |- changelog.rst
|   |  ...
|   |- conf.py
|   |- make.bat
|   |- Makefile
|   |- _static/*
|   \- _templates/*
\- src/
    \- <package>
        |- __init__.py
        |- __main__.py
        |- __meta__.py
        |- <package>.py
        \- tests/
            |- __init__.py
            |- __main__.py
            \- test_<package>.py

The main entry point for installation and management of your project. We provide a minimum working script based on setuptools. In order to avoid duplication of work, the setup.py script reuses the project description from README.rst and uses the version string in a separate VERSION file. Users still have to edit this file and provide the short description of the project.

Don't forget to python setup.py register your project to PyPI before someone takes it name!

python-boilerplate puts all source code for your project under the src/<package> folder. This contrasts with the other typical approach of leaving the python packages directly in the root of the the source tree. We believe that a separate src folder is more organized and manageable in the long run.

We also create a "<package>.tests" module for unit testing and distribute it with the main package. The drawback of this approach is a slightly larger distribution. In most systems, this small price is greatly offset by the ability to ask users to easily run the test suite when dealing with bug reports. python-boilerplate creates a __main__.py file in the tests package that enable anyone can run the test suite simply by calling python -m <package>.tests.

python-boilerplate creates the skeleton for a Sphinx-based documentation. The documentation reuses both the README.rst and INSTALL.rst files. In most cases, it is probably a good idea to create a relatively small README.rst with a succinct overview of your project and leave most details of the documentation to specific files inside the docs/ directory.

The README.rst file in python-boilerplate itself is perhaps too big ;-)

_ Sphinx: https://sphinx-doc.org

We provide a default INSTALL.rst file with generic installation instructions for Python packages. Unless your project requires something fancy, this probably can be left as is.

The README.rst file, however, provides a detailed overview of your project. You should edit this file to provide a meaningful description, otherwise a not so flattering default will be used. The contents of README.rst are also displayed in the index page of the project's documentation.

Your project's version is conveniently centralized in a single file. The setup.py script uses this value to register you package and it also saves the correct version in the <package>.__version__ attribute in your module.

You may bump version numbers using an invoke task:

$ inv bump-version

This method assumes that the version string is in the form "<major>.<minor>.<micro>".

The requirements.txt uses the - e . directive to tell pip to search for the requirements in the setup.py script. As a general rule, dependencies should be specified only in the install_requires flag in your setup.py.

You may want to use your requirements.txt to freeze packages to specific versions by adding lines such as:

my-package==1.2.3

Freezing makes sense for packages that are meant to run only on their own private environments such as a Django project running in it own virtualenv or docker container. Avoid freezing package versions in your main Python installation.

Define files to be included in the source distributions created by setuptools.

Python boilerplate accepts the most common open source licenses (or at least it should). If the license you want to use is not supported, we gladly accept patches!

The default .gitignore excludes python bytecode and all build directories.

The tasks.py define some invoke tasks for your project. You can define new tasks by defining python functions just as the example given in this file. Think of tasks.py as a Python replacement of a Makefile: it is used to define commands that automate repetitive tasks and chores. We define a few general purpose tasks. They are executed using the inv(oke) command.

inv test:

Runs py.test with the main test suite.

inv coverage:

Runs py.test and display a coverage report.

inv build:

Calls setup.py build and also builds the documentation.

inv bump-version:

Controls the version number in the VERSION file.

python-boilerplate ships a working .travis.yml file and a tox.ini. You can use tox to run the test suite for different Python versions locally (but you'll need several working interpreters simultaneously installed in your system).

Assuming that you are hosting your code at Github, enable Travis-CI integration under "Settings > Integrations and services" option in your main repository page. Also enable "Coveralls" integration to have good quality reports on code coverage evolution.

You need to enable support for your repository both in Travis-CI<https://travis-ci.org> and Coveralls<http://coveralls.io> websites. Continuous integration tasks will run every time you push something new to Github.