The cliff framework is meant to be used to create multi-level commands such as subversion and git, where the main program handles some basic argument parsing and then invokes a sub-command to do the work.
Cliff takes advantage of Python’s ability to load code dynamically to allow the sub-commands of a main program to be implemented, packaged, and distributed separately from the main program. This organization provides a unified view of the command for users, while giving developers the opportunity organize source code in any way they see fit.
Cliff is organized around four objects that are combined to create a useful command line program.
An cliff.app.App is the main program that you run from the shell command prompt. It is responsible for global operations that apply to all of the commands, such as configuring logging and setting up I/O streams.
The cliff.commandmanager.CommandManager knows how to load individual command plugins. The default implementation uses setuptools entry points but any mechanism for loading commands can be used by replacing the default CommandManager when instantiating an App.
The cliff.command.Command class is where the real work happens. The rest of the framework is present to help the user discover the command plugins and invoke them, and to provide runtime support for those plugins. Each Command subclass is responsible for taking action based on instructions from the user. It defines its own local argument parser (usually using argparse) and a take_action() method that does the appropriate work.
The main program uses an cliff.interactive.InteractiveApp instance to provide a command-shell mode in which the user can type multiple commands before the program exits. Many cliff-based applications will be able to use the default implementation of InteractiveApp without subclassing it.