Skip to content

Python Guidelines

Jump to bottom Edit New page
Ned Batchelder edited this page Oct 6, 2015 · 27 revisions

General Python

Generally, don't edit files just to change the style. But do aim for this style with new or modified code.

See also http://edx-developer-guide.readthedocs.org/en/latest/testing/code-quality.html

Follow PEP 8

  • 4-space indents (no tabs)
  • Names like this: modules_and_packages, functions_and_methods, local_variables, GLOBALS, CONSTANTS, MultiWordClasses
  • Acronyms should count as one word: RobustHtmlParser, not RobustHTMLParser
  • Trailing commas are good: they prevent having to edit the last line in a list when adding a new last line. You can use them in lists, dicts, function calls, etc.
  • EXCEPT: we aren't (yet) limiting code lines to 79 characters. Use 120 as a limit for code. Please use 79 chars as a limit for docstring lines though, so that the text remains readable.

Breaking long lines

Follow these guidelines:

  • Try to refactor the code to not need such long lines. This is often the best option, and is often overlooked for some reason. More, shorter lines are good.
  • If you need to break a function call over more than one line, put a newline after the open paren, and move the arguments to their own line. DO NOT indent everything to where the open paren is. This makes the code too indented, and makes different function calls near each other indented different amounts.

# NO NO NO!:

results = my_object.some_method(arg1,    # this is very
                                arg2,    # very ugly and makes
                                arg3,    # the code squished over on the right.
                                )

# YES:

results = my_object.some_method(
    arg1,
    arg2,
    arg3,
)

# OR:

results = my_object.some_method(
    arg1, arg2, arg3
)

Important points:

  • Don't over-indent to make things line up with punctuation on the first line.
  • Closing paren should be on a line by itself, indented the same as the first line.
  • The first line ends with the open paren.

Imports Order

PEP 8 recommends a most-general to most-specific import order, which means this order:

  • Standard library imports
  • Third Party Library imports
  • Core Django imports
  • Third party Django app imports
  • Other edX repo imports
  • Local imports

Follow PEP 257

  • Write docstrings for all modules, classes, and functions.
  • Use three double-quotes for all docstrings.
  • Start with a one-line summary. If you can't fit a summary in one line, think harder, or refactor the code.
  • Write in Sphinx-friendly prose style. Put backquotes around code names (variables, parameters, methods, etc).

The preferred style is so-called "Google Style" with readable headers for different sections, and all arguments and return values defined.

Note: there is one exception. REST APIs created using Django REST Framework (DRF) must use a hybrid format that is suitable both for DRF and ReadTheDocs. This format is documented here: https://openedx.atlassian.net/wiki/display/AC/edX+REST+API+Conventions#edXRESTAPIConventions-docstringsDocstringFormat

See:

Here's how you write documentation in "Google Style":

def func(arg1, arg2):
    """Summary line.

    Extended description of function.

    Arguments:
        arg1 (int): Description of arg1
        arg2 (str): Description of arg2

    Returns:
        bool: Description of return value

    """

Most of our code is written using an older style:

def calculate_grade(course, student):
    """
    Sum up the grade for a student in a particular course.

    Navigates the entire course, adding up the student's grades. Note that
    blah blah blah, and also beware that blah blah blah.

    `course` is an `EdxCourseThingy`. The student must be registered in the
    course, or a `NotRegistered` exception will be raised.

    `student` is an `EdxStudentThingy`.

    Returns a dict with two keys: `total` is a float, the student's total
    score, and `outof` is the maximum possible score.

    """

If you only have a single line in your docstring, first, consider that this is almost certainly not enough documentation, and write some more. But if you do have just one line, format it similarly to a multi-line docstring:

def foo(a, b):
    """
    Computes the foo of a and b.
    """

Not like this:

def foo(a, b):
    """Computes the foo of a and b.""" # NO NO NO

Write a Good repr() for Each Class

Each class should have a __repr__() method defined, so that calling repr() on an instance of the class returns something meaningful that distinguishes objects from each other to a human being. This is useful for debugging purposes.

Pylint Guidelines and Practices

  • For unused args, you can prefix the arguments with _ to mark them as unused (as convention), and pylint will accept that.

  • Adding a TODO in one place requires you to make a pylint fix in another (just to force us to clean up more code)

  • No bare except clauses. except: should be except Exception:, which will prevent it from catching system-exiting exceptions, which we probably shouldn't be doing anyway. If we need to, we can catch BaseException (There's no point in catching BaseException, that includes the exceptions we didn't want to catch with except: in the first place.) (ref: http://docs.python.org/2/library/exceptions.html#bltin-exceptions). Catching Exception, however, will still generate a pylint warning "W0703: catching too general exception." If you still feel that catching Exception is justified, silence the pylint warning with a pragma: "# pylint: disable=broad-except"

  • Although we try to be vigilant and resolve all quality violations, some Pylint violations are just too challenging to resolve, so we opt to ignore them via use of a pragma. A pragma tells Pylint to ignore the violation in the given line. An example is:

    self.assertEquals(msg, form._errors['course_id'][0])  # pylint: disable=protected-access

The pragma starts with a # two spaces after the end of the line. We prefer that you use the full name of the error (pylint: disable=unused-argument as opposed to pylint: disable=W0613), so it's more clear what you're disabling in the line.

Classes vs Dictionaries

It's better to use a class or a namedtuple to pass around data that has a fixed shape than to use a dict. It makes it easier to debug (because there is a fixed, named set of attributes), and it helps prevent accidental errors of either setting new attributes into the dictionary (which might, for instance, get serialized unexpectedly), or might be typos.

Django Good Practices

Imports

Always import from the root of the project:

from lms.djangoapps.hologram.models import 3DExam    # GOOD
from .models import 3DExam                           # GOOD
from hologram.models import 3DExam                   # BAD!

The second form (relative import) only works correctly if the importing module is itself imported correctly. As long as there are no instances of the third form, everything should work. Don't forget that there are other places that mention import paths:

url(r'^api/3d/', include('lms.djangoapps.hologram.api_urls')),   # GOOD
url(r'^api/3d/', include('hologram.api_urls')),                  # BAD!

@patch('lms.djangoapps.hologram.models.Key', new=MockKey)        # GOOD
@patch('hologram.models.Key', new=MockKey)                       # BAD!

Design

  • Fat Models, Helper Modules, Thin Views, Stupid Templates

References

Add a custom footer
Add a custom sidebar
Clone this wiki locally