Skip to content
This repository has been archived by the owner on Apr 7, 2022. It is now read-only.

Developer Guidelines

Jump to bottom
psav edited this page Dec 18, 2013 · 15 revisions

General Guidelines

Contributing

  • Own your pull requests; you are their advocate.
    • If a request goes unreviewed for two or three days, ping a reviewer to see what's holding things up.
    • Follow up on open pull requests and respond to any comments or questions a reviewer might have.
  • Keep the contents of the pull request focused on one idea. Smaller pull requests are easier to review, and thus will be merged in more quickly.
  • After submitting a request, be ready to work closely with a reviewer to get it tested and integrated into the overall test suite.
  • Follow the Code Style guidelines to make your pull request as easy to review as possible.
  • If your request requires the use of private information that can't be represented in the data file templates (probably cfme_data.yaml), please state that in the test module docstring or the individual test docstring, along with information on where that data can be found.
  • Similar to the last point, any data files used by a test module should be clearly documented in that module's docstring.
  • Any data required in a sensitive data file should be reflected in the template for that file.
  • Standards may change over time, so copying older code with similar functionality may not be the most productive action. If in doubt, refer back to this document and update the copied code according to the current guidelines.
  • Please keep large lint changes separate from new features, though this point should become less relevant over time
  • All pull requests should be squashed down to logical blocks of distinctive functionality that work by themselves and do not result in brokenness of master
    • As an example, if you were working on a test which required new pages, utilities and tests, it would be OK to split the page, utility and test changes into separate requests or commits, providing they were in the correct order of dependency.

Reviewers

Reviewers will be looking to make sure that the Contributing guidelines are being met. Some of the things that go into the review process:

  • Assign the PR to the reviewer
  • Pull request branches will be rebased against current master before testing.
  • Newly added tests will be run against a clean appliance.
  • Adherence to code style guidelines will be checked.

If tests fail, reviewers WILL:

  • ...give you a complete traceback of the error.
  • ...give you useful information about the appliance against which tests were run, such as the appliance version.
  • ...give you insight into any related data files used.

If tests fail, reviewers WILL NOT:

  • ...thoroughly debug the failing test(s).

All requests require 2 approvals from two reviewers, after which time, the contributor may, permissions allowing, merge the commit him/herself.

Reviewers must never approve their own pull requests.

Code Style

We adhere to Python's PEP 8 style guide, occasionally allowing exceptions for the sake of readability. This is covered in the "Foolish Consistency" section of PEP 8. Information on using linting tools to help with this can be found on the linty freshness page.

We also do a few things that aren't explicitly called out in PEP 8:

  • The github pull request pane is our primary code review medium, and has a minimum width of 100 characters. As a result, our maximum line length is 100 characters, rather than 80.

  • When wrapping blocks of long lines, indent the trailing lines once, instead of indenting to the opening bracket. This helps when there are large blocks of long lines, to preserve some readability:

      _really_really_long_locator_name = (True, "div > tr > td > a[title="this \
      is just a little too long"])
  _another_really_super_long_locator_name = (True, 'div > tr > td > \
      a[title="this is getting silly now'])

  • When wrapping long conditionals, indent trailing lines twice, just like with function names and any other block statement (they usually end with colons):

      if this_extremely_long_variable_name_takes_up_the_whole_line and \
          you_need_to_wrap_your_conditional_to_the_next_line:
      # Two indents help clearly separate the wrapped conditional
      # from the following code.

  • When indenting a wrapping sequence, one indent will do. Don't try to align all of the sequence items at an arbitrary column:

      a_good_list = [
      'item1',
      'item2',
      'item3'
  ]
      a_less_good_list = [ 'item1',
                       'item2',
                       'item3'
  ]

  • According to PEP 8, triple-quoted docstrings use double quotes. To help differentiate docstrings from normal multi-line strings, consider using single-quotes in the latter case:

      """This is a docstring.

  It follows PEP 8's docstring guidelines.

  """
      paragraph = '''This is a triple-quoted string, with newlines captured.
  PEP 8 and PEP 257 guidelines don't apply to this. Using single quotes here
  makes it simple for a reviewer to know that docstring style doesn't apply
  to this text block.'''

  • On the subject of docstrings (as well as comments) use them. Python is somewhat self-documenting, so use docstrings and comments as a way to explain not just what code is doing, but why it's doing what it is, and what it's intended to achieve.

    We have decided to use the following docstring format and use the "Cartouche" Sphinx plugin to generate nice docs. Details on the format can be found above, but an example is described below.

      def my_function(self, locator):
      '''Runs the super cool function on a locator

      Seriously, you have to try this

      Note: You don't actually have to try it

      Args:
          locator: The name of a locator that can be described by using
              multiple lines.

  Returns:
          Nothing at all.

      Raises:
      CertainQuestionsError: Raises certain questions about the authors sanity.

  • In addition to being broken up into the three sections of standard library, third-party, and the local application, imports should be sorted alphabetically. 'import' lines within those sections still come before 'from ... import' lines:

      import sys
  from os import environ
  from random import choice

Other useful code style guidelines:

cfme_tests

Layout

cfme_tests/

  • pages/ Top-level pages container. The structure of this directory should mimic the layout of the CFME UI as much as possible.
  • tests/ Top-level tests container
    • appliance/ Appliance tests, generally using tools other than the UI to inspect appliance internals, like verify installed packages or required service states
    • scenario/ Large test scenarios, representing complicated setup and teardown workflows with interdependent tests
    • ui/ General UI tests, testing specific functional units of behavior
  • data/ Test data. The structure of this directory should match the structure under tests/, with data files for tests in the same relative location as the test itself.
    • For example, data files for tests/ui/test_ui_widgets.py could go into data/ui/test_ui_widgets/.
  • fixtures/ py.test fixtures that can be used by any test. Modules in this directory will be auto loaded.
  • markers/ py.test markers that can be used by any test. Modules in this directory will be auto loaded.
  • utils/ Utility functions that can be called inside our outside the test context. Generally, util functions benefit from having a related test fixture that exposes the utility to the tests. Modules in this directory will be auto loaded.
  • db/ The 'db' module provides access to an appliance's database, as well as convenience mappings for model to table names.
  • scripts/ Useful scripts for QE developers that aren't used during a test run

General Notes

  • Avoid using time.sleep as much as possible to workaround quirks in the UI. There is a wait_for utility that can be used to wait for arbitrary conditions. In most cases there is some DOM visible change on the page which can be waited for.
  • Avoid using time.sleep for waiting for changes to happen outside of the UI. Consider using tools like mgmt_system to probe the external systems for conditions for example.
  • If you feel icky about something you've written but don't know how to make it better, ask someone. It's better to have it fixed before submitting it as a pull request ;)

Writing Pages

Pages are read-only python modeling of the CFME UI, used by tests

Pages to model the CFME UI in Python, allowing the functional tests of the UI to be ignorant of the underlying page structure. As such, UI elements (pages, regions, forms, etc.) modeled in cfme_pages must provide helper methods and properties to expose a usable interface to cfme_tests. This is explained in more detail in the section on "Writing Tests".

As mentioned in the README, pages should be modeled as a part of writing tests. Code in cfme_pages must never depend on code in cfme_tests.

When writing pages, a few points should be noted:

  • Follow the standard naming convention for locators
    • Name of element, followed by type from the type list
  • Type list: button, select, text, textbox, radio, option
  • Ensure that your element is presented in an expected way. As an example, presenting a div containing and unordered list in one place and simply the unordered list in another, causes an unpredictable inconsistency as to how to handle the locator.
  • Try to avoid using localized text as part of a locator where possible

Writing Tests

Tests in cfme_tests have the following properties:

  • They pass on a freshly deployed appliance with no configuration beyond the defaults (i.e. tests do their own setup and teardown).
  • They never directly access a page's 'testsetup' attribute, or call selenium methods. Instead, methods defined on the page objects will carry out the required behavior.
  • Where possible, they strive to be idempotent to facilitate repeated testing and debugging of failing tests. (Repeatable is Reportable)
  • Where possible, they try to clean up behind themselves. This not only helps with idempotency, but testing all of the CRUD interactions helps to make a thorough test.
  • Tests should be thoroughly distrustful of the appliance, and measure an action's success in as many ways as possible. A practical example:
    • Do not trust flash messages, as they sometimes tell lies (or at least appear to). If you can go beyond a flash message to verify a test action, do so.

Some points when writing tests

  • When naming a test, do not use a common part of multiple test names as a test name itself. In the example below, trying to run a single test called test_provider_add, not only runs that test, but also test_provider_add_new and test_provider_add_delete, as pytest uses string matching for test names. test_provider_add should have a suffix making it unique. In this way a tester can choose the run just the single test on its own, or the group of tests, whose names all begin the same way.
    • test_provider_add - Adds a provider
    • test_provider_add_new - Adds a new provider type
    • test_provider_add_delete - Adds a provider and then deletes it
  • Where a clean-up is required, it should be carried out in a Finalizer. In this way we prevent leaving an appliance dirty if the test fails as the clean up will happen regardless.
  • Keep all properties, fixtures and functions together

Fixtures

Fixtures are not only responsible for setting up tests, but also cleaning up after a test run, whether that test run succeeded or failed. addfinalizer is very powerful. finalizer functions are called even if tests fail.

When writing fixtures, consider how useful they might be for the overall project, and place them accordingly. Putting fixtures into a test module is rarely the best solution. Instead, try to put them in the nearest conftest.py. If they're generic/useful enough consider putting them into one of the fixtures/ directory for use in cfme_tests or the plugin/ directory for use in both projects.

This Document

This page is subject to change as our needs and policies evolve. Suggestions are always welcome.

Clone this wiki locally