The goal of this document is to establish a common understanding among software contributors to D-Wave’s Ocean software projects based on the code conventions and best practices used at D-Wave.
This document is intended to be a living document, and is not complete. Feedback is welcome.
A feature or bugfix is complete only when the code has been unit-tested. Part of the pull-request process is to test your branch against master, and so the more tests you provide, the easier the pull-request process.
When contributing to a D-Wave project, fork the repository using the Github fork button and work in a feature branch in your own repository. In your feature branch, commit and push often, you can always rebase locally to edit your commit history and make it readable. To start a discussion, initiate a pull request. The sooner you start the pull request, the better. This allows early discussion on your feature and saves time and effort when you are ready to merge. When you are ready to merge, notify the repository owner and they will merge your code in.
Follow the commit conventions described here:
Separate subject from body with a blank line
Limit the subject line to 50 characters
Capitalize the subject line
Do not end the subject line with a period
Use the imperative mood in the subject line
Wrap the body at 72 characters
Use the body to explain what and why vs. how
If your branch is long-lived, rebase off of master periodically.
If you’re already familiar with Git, but rebasing is still scary, try this think like a Git guide.
The master branch in a dwavesystems repository reflects the bleeding-edge developments of the project. Significant old versions forked from master are kept in version branches. For example, we might keep version branches 2.x and 3.x while master tracks the latest 4.x version. We try to minimize the number of version branches kept alive/maintained.
Stable releases are tracked using tags. These tagged snapshots are deployed to PyPI after successfully passing the test cases during continuous integration.
Variable naming should follow the well-known conventions of a language. Avoid uninformative or needlessly terse variable names.
Code is read more often than written.
Functions should do one thing.
Early pull requests and code reviews.
Early architecting/design. Code reviews can happen before any code has been written.
Use a consistent character width of 120.
Use 4 spaces instead of tabs.
End all files with a newline.
Documentation and Comments#
Do a good job of commenting. Read this Coding Horror article.
Comments should add, not repeat: avoid repeating in English or pseudo-code what the code does. Rather, discuss what the block is trying to achieve.
Side effects should be visible on screen; if not in code, then in comments.
Remember, the best documentation is clean, simple code with good variable names. When this is not possible, you must use a comment to explain the purpose of a functional block.
The following code
# z must not be greater than 255. if z > 255: raise RuntimeError('z must be <= 255!')
would be much more informative as
# if z is greater than 255, this universe will collapse. See https://url.to.issue.tracker/IS-42 if z > 255: raise RuntimeError('z must be <= 255!')
or even better:
# See https://url.to.issue.tracker/IS-42 if z > 255: raise RuntimeError('z cannot be greater than 255, or this universe will collapse.')
As a baseline, follow the pep8 style guide for python.
Private functions should include some sort of docstring.
If your module has more than one public unit, it should have a module docstring with a table of contents.
The docstring for the
__init__method goes on the class.
C++ code should be compatible with standard C++11.
Our format is based on Google C++ style guide with some exceptions:
Column width is limited to 120 characters. Best effort should be made to keep to 80 characters, but up to 120 can be used for clarity.
The base indent level is 4.
Non-const references are allowed.
When starting a new C++ project, use clang-format with the .clang-format file included here.
Our code follows Semantic Versioning conventions: major.minor.patch.
A change that breaks backwards compatibility must increment the major version. Anything below version 1.0.0 can break backwards compatibility.
If you are creating a repository, don’t forget to include a
a reasonable description of your project.