How should I document a Python script?

Question

I have a couple of Python modules that are meant to be run as scripts. How should I write the docstrings at the module and function level to make it clear how to run and use the module?

• Thomas Owens: Are these scripts run by users manually or automatically as part of some process? Who is the intended user?
• murgatroid99: One of the scripts is run as part of a cron job, but all of them have to be usable by users.

Answer 1

Personally I use the docstrings for other programmers/installers, there I document the purpose and any setup details that you need to look at rarely. For user documentation, I use the argparse module so users can run: program --help

import argparse
# other stuff
parser = argparse.ArgumentParser(
 description='Path building utility.',
 epilog="""Default operation is to read a list of directory
 names from stdin and print a semi-colon delimited
 path list to stdout. To terminate --add, --append,
 --prepend or --remove list, use -- as in: {}
 --remove *foo* *bar* -- input.file
 """.format(os.path.basename(sys.argv[0])))
parser.add_argument('--add', '--prepend', 
 help='add specified directories to the front of the path', 
 dest='front', type=str, nargs='+', metavar='DIR', action='append')
# etc.

• murgatroid99: This looks like a good idea, but at the moment my scripts have few enough options that it would probably add some unnecessary complexity.

Answer 2

I'd mimick a man(1) page

NAME
 this command does things
SYNOPSIS
 command do-things nice
DESCRIPTION
 things are nice when this command does them

Answer 3

In addition to ZJR's suggestion of mimicking a man page, you can also provide a usage statement, return codes, and in-code documentation.

I would consider providing a usage statement that explains how to invoke the application in a clear, concise manner. This would simply provide a list of the arguments, what the arguments represent, if a given argument is optional, and any other relevant information. This would be convenient to users who invoke your scripts from the command-line.

For automated applications that call your scripts, I would consider using various return codes to document errors or other non-standard execution considerations. These can be easily parsed and would shift handling of exceptional cases (especially those that you simply can't handle in your script) as well as logging of the event into the calling application.

Using docstrings might be overkill, depending on the complexity of the script, but some form of in-code documentation is needed for future developers (including youself in the future). I would include a high-level overview of what the script does as the very first thing in the file. Self-documenting code would always help, as well, so make sure you've got good variable names. If there's any complex logic, make sure it's clear what's going on and why you made the design decisions that you did with comments.