At it simplest, a Python package is a mere directory with a '__init__.py' file in it. However, this basic structure needs to be augmented as soon as more functionality is required: i.e. create a distribution, write a comprehensive documentation, run some tests. With time the structure of a package grows and include more and more description files (e.g. setup.py, .gitignore, ...).

The rationale behind the creation of this 'package builder' is to keep the life of a python programmer as easy as possible by providing two core functions:

• a way to add more functionality to an existing package
• a way to keep the package structure up to date with currently known best practices.

Create a virtual environment for development:

$ virtualenv dvlpt

Activate it:

$ dvlpt/Scripts/activate

Install ltpkgbuilder:

(dvlpt)$ pip install ltpkgbuilder

Create a directory for your package:

(dvlpt)$ mkdir toto

Run 'manage' inside this directory:

(dvlpt)$ cd toto
(dvlpt)toto$ manage init
(dvlpt)toto$ manage add -opt base
(dvlpt)toto$ manage regenerate

This will create the bare basic minimum for a python package. Add more options (see Add Package Structure Functionality for more options) afterward. Especially, since in the example above we just added the 'base' option that will create a 'src' directory to put your code in it.

Packages generated with Package Builder contains three different types of files:

• 'pkg_cfg.json', a resource file that contains the information you entered at some stage during the configuration phase of adding an option.
• generated files, susceptible to be regenerated at any time or with version change and not meant to be modified by user. These files are generated automatically by the package builder using templates provided with the package.
• developer data and modules edited by hand which contains the actual python code of the package independently of the structure of the package. ltpkgbuilder will never touch them. If they conflict with some files used by a newly added option, the user will be prompted and will have to solve the conflict to install the option.

A call to the 'update' command will check for new versions of the package or any available option:

(dvlpt)toto$ manage update

This command requires an internet connection since local installation will be compared to current code on github.

If a newer version exists, you will be prompted for installation. After a successful installation you will be prompted for new arguments if the configuration of one of your installed options was upgraded in the process.

If update is successful, a call to regenerate is mandatory to rebuilt the package structural files:

(dvlpt)toto$ manage regenerate

This phase will never overwrite any files you modified or created. You'll be prompted in case of conflicts but it is your responsibility to solve them and relaunch the command.

Package Builder provide a set of options to introduce new functionality to an already existing package:

• base: base option, basic package management
• license: will help the developer to choose a license and add the relevant files
• doc: Add some documentation to your package
• test: basic unitests using Nose
• coverage: add code coverage to the basic test configuration
• pydist: make your package distributable with setuptools (i.e. setup.py)
• data: will guide through all the steps to add non python files to a package
• github: will guide through all the step to safely store the package on Github
• readthedocs: step by step guide to have your documentation on ReadTheDocs
• travis: will guide through all the steps to compile the code on Travis-CI
• tox: defines config files to use multi environment testing, Tox
• flake8: install and config Flake8 tools to check for code compliance to PEP8 guidelines.
• pypi: step by step guide and configuration to upload package on PyPi.

To install a new option call the 'add' action:

(dvlpt)toto$ manage add -opt license

The script will perform different tasks sequentially:

• Check if this option requires other options in order to be installed: e.g. the 'pydist' option requires all 'base, 'doc', 'test', 'license' and 'version' in order to run properly.
• Check if this option requires some extra packages in order to setup: e.g. the 'license' option depends on the lice package to function properly.
• Run a basic config script to ask you for specific details relative to this option e.g. the 'license' option will ask for the license name.

Multiple call to add options can be serialized but you explicitly needs to call regenerate to see the action of the new options on your package:

(dvlpt)toto$ manage regenerate

Each option comes with some example files that can be installed with the special directive:

(dvlpt)toto$ manage add -opt example

You will be prompted for the name of the option of the files you want to install.

The files will be directly installed without the need to a regenerate call. They have a special status in the sense that you can modify or even remove these files without any complains next time you rebuild the package. You can also reinstall them at any time (you'll be prompted for action if conflicts occur).

(dvlpt)toto$ manage add -opt example -e option_name base

You can simply edit an option (e.g. license) by running the command:

(dvlpt)toto$ manage edit -opt license

You'll be re-prompted for the values of arguments of this option with default to previously entered values.

Package Builder also provides a few useful services to check that the python modules follow code best practices:

• 'add_object': will create a new python module with the proper headers and a skeleton of a python class.
• 'add_plugin': will wrap a given python class into a usable plugin.
• 'add_script': will wrap a given python functionality into a command line script.
• 'reset_file_header': will loop through all python modules and try to rewrite file header to match current best practices.
• fmt_doc: check code documentation and format it according to given standard if possible. Requires some already good documentation, just a quick fix to pass from one style to another (e.g. google to numpy).

You can contribute to this package by:

• improving the documentation
• correcting some bugs
• closing a few issues
• implementing a new option to add a new functionality to package structures