Modern Python Documentation Standards

I'm not going to focus much on the syntax. Instead I rather explain principles I used to guide me. Python is 30 years old and there are artifacts of the language should probably go away.

"Code is more often read than written." - Guido Van Rossum, Creator of Python

When I learn and distill information, it's helpful for me to write a one sentence description and to category when and where to use things.

What is the purpose of each type of doc element?

  • Doctests: Allows you to embed testable code snippets which seems essential in the early days of Python. Stylistically, this is not a great way to document, so use it sparingly.
  • Docstrings: Should always be use for public classes, modules, and functions. Not required for non-public but you should describe each method.
  • NumPy style docstrings: Provides a structured and highly verbose style to document multi-paragraph annotations, parameters, return values, and links to documentation. For me this is a super class of docstrings. Suggested for AI and ML projects, open source projects, and scientific applications.
  • Block comments - generally apply to some or all of the code that follows.
  • Inline Comments
  • Tagging Comments: Adding comments that are suggestions or meant to imply future work: #BUG, #TODO, #Deprecated: remove in the future.

Documenting vs. Commenting

Code Comments

When code is concise and well structured, reading code tells you how it works.

Comments should explain why you've written this code.  

  • Comments describe code for other developers
  • Makes extending and maintaining the code easier.
  • Helps the reader build a mental model of the code by describing the purpose and design.
# What business rule you implemented?
def display_name():
	# Policy to use title and full name of customer.
    return "President Abraham Lincoln"
# Why did you made this decision?
def utc_now():
	# All datetime values must be TZ aware and set UTC timezone.

def expensive_computation():
    # Creates a small in memory cache to reused O(n^3) computation 

Code Documentation

  • Describes code use and functionality to users
  • May aid further development of code, but intended audience is users (both actual and potential) of the code.

A single line of code can be just as inscrutable as a thousand lines of code. Bad documentation is just as harmful as no documentation. There will never be a perfect or scientifically provable method for creating software.

It's on us to accept that most aspects of software development is subjective. Errors exist.  There is no right way to program.

Show Comments