Development Team/Code guidelines
The community around Sugar is very diverse and only a small part of it is directly involved in software development. Of those, an even smaller part is personally responsible for the quality of the software, its growth, and its availability to translators, distributors and deployers. See Development Team/Release/Modules for a list of modules with their maintainers and peers.
But this smaller group of people aren't the only ones that can contribute to the development of Sugar itself. Anyone with some spare time and enough skills and passion can make significant contributions to Sugar, but in order to bound to a limit the maintenance work, the submitted code needs to have some characteristics, explained in this page.
Please take these recommendations with a grain of salt, they are intended to make the life of your fellow developers easier, and are not intended to be general rules of "good programming".
It's all about communication
Your code will run as interpreted by the Python interpreter, but other people won't be able to efficiently work with it if you don't address people as well when writing it. Write your code first for people and secondarily for the machine.
If you are going to do things differently, be sure to have a really good reason. Often a bad standard is better than no standard at all.
Just because it looks smarter is probably a bad reason, as most of the people that will be reading at your code won't be doing it for pleasure or edification, but in order to fix bugs or add features. In those cases, it's better when code is boring and predictable.
Avoid reference cycles, and use signals to break such cycles when redistributing responsibility is not enough.
Keep the size of your modules, classes, methods and functions under reasonable limits. Splitting a method in two is a good opportunity to put a name to a block of code.
Make sure that API policy rules are respected.
First rule is to use the same style as the existing code around your new method, class or module. Sugar code is quite consistent as of now (August 6th 2009) so you shouldn't need to change styles often except when modifying very old code.
Follow PEP 08, but as it is quite broad, it's not enough in itself to guarantee a minimum of consistency. Some more comments below.
Use inline comments sparingly. If you feel your code needs an inline comment, consider instead splitting the method in two, with a reasonable name and API comment. API comments should say what the code does, not how it does it. How your code works should be clear from the code structure, refactor until the code structure is a clear reflection of its purpose.
Python is a very expressive language that allows for different programming models, before using the more exotic features consider readability over compactness and think that we want people to be able to contribute to Sugar without needing to be familiar with the whole Python syntax.
- Use logging.debug('foo %r bar', x) instead of logging.debug('foo %r bar' % x). This is more robust and formatting only happens if we actually log something.
- When you catch an exception and want to log it, use logging.exception(), this will print the whole backtrace and exception information which is very useful when debugging.
When possible use tools to check your code, this will save lots of time for everybody involved.
Please try to use pylint to verify your patch for things like exceeding 80 columns etc., unused imports and unused variables. Pylint is not a tool you can rely on 100% but it helps to follow some guidelines and to avoid the most stupid errors like typos. sugar-jhbuild contains a pylintrc you can use in the
scripts/data directory: pylintrc (follow "raw blob data").
pep8.py catches more style errors than pylint, so make sure to run that one, too.
In the sugar packages use 'make distcheck' to make sure all files are included and the POTFILES.in is up to date.
We are running periodic builds of Sugar using Buildbot. If you break the build you are responsible to get it fixed. If you don't, the release team will take care of it, most likely by reverting your patch. More in detail:
- Every change that causes build or check failures should be immediately fixed or reverted.
- Every change that causes check warnings should be fixed or reverted within 24 hours.
You can use the check command in jhbuild to execute the same checks that Buildbot is doing.
Most projects have their own style guides, it's often useful to read them and understand the rationales behind the non-arbitrary bits.