Skip to content
OpenStack Hacking Style Checks
Python Shell
Latest commit 7d5947b Apr 21, 2016 OpenStack Proposal Bot Updated from global requirements
Change-Id: Ib6b65007eda14982f8f3cdb62823e2e850982cd5

README.rst

Introduction

hacking is a set of flake8 plugins that test and enforce the :ref:`StyleGuide`.

Hacking pins its dependencies, as a new release of some dependency can break hacking based gating jobs. This is because new versions of dependencies can introduce new rules, or make existing rules stricter.

Installation

hacking is available from pypi, so just run:

pip install hacking

This will install specific versions of flake8 with the hacking, pep8, mccabe and pyflakes plugins.

Origin

Hacking started its life out as a text file in Nova's first commit. It was initially based on the Google Python Style Guide, and over time more OpenStack specific rules were added. Hacking serves several purposes:

  1. Agree on a common style guide so reviews don't get bogged down on style nit picks. (example: docstring guidelines)
  2. Make code written by many different authors easier to read by making the style more uniform. (example: unix vs windows newlines)
  3. Call out dangerous patterns and avoid them. (example: shadowing built-in or reserved words)

Initially the hacking style guide was enforced manually by reviewers, but this was a big waste of time so hacking, the tool, was born to automate the process and remove the extra burden from human reviewers.

Versioning

hacking uses the major.minor.maintenance release notation, where maintenance releases cannot contain new checks. This way projects can gate on hacking by pinning on the major.minor number while accepting maintenance updates without being concerned that a new version will break the gate with a new check.

For example a project can depend on hacking>=0.10.0,<0.11.0, and can know that 0.10.1 will not fail in places where 0.10.0 passed.

Adding additional checks

Each check is a pep8 plugin so read

The focus of new or changed rules should be to do one of the following

  • Substantially increase the reviewability of the code (eg: H301, H303) as they make it easy to understand where symbols come from)
  • Catch a common programming error that may arise in the future (H201)
  • Prevent a situation that would 100% of the time be -1ed by developers (H903)

But, as always, remember that these are Guidelines. Treat them as such. There are always times for exceptions. All new rules should support noqa.

If a check needs to be staged in, or it does not apply to every project or its branch, it can be added as off by default.

Requirements

  • The check must already have community support. We do not want to dictate style, only enforce it.
  • The canonical source of the OpenStack Style Guidelines is :ref:`StyleGuide`, and hacking just enforces them; so when adding a new check, it must be in HACKING.rst
  • False negatives are ok, but false positives are not
  • Cannot be project specific, project specific checks should be Local Checks
  • Include extensive tests
  • Registered as entry_points in setup.cfg
  • Error code must be in the relevant Hxxx group
  • The check should not attempt to import modules from the code being checked. Importing random modules, has caused all kinds of trouble for us in the past.

Local Checks

hacking supports having local changes in a source tree. They can be configured to run in two different ways. They can be registered individually, or with a factory function.

For individual registration, put a comma separated list of pep8 compatible check functions into the hacking section of tox.ini. E.g.:

[hacking]
local-check = nova.tests.hacking.bad_code_is_terrible

Alternately, you can specify the location of a callable that will be called at registration time and will be passed the registration function. The callable should expect to call the passed in function on everything if wants to register. Such as:

[hacking]
local-check-factory = nova.tests.hacking.factory
Something went wrong with that request. Please try again.