Disallowed Minor Code Changes

There are a few types of code changes that have been proposed recently that have been rejected by the Glance team, so we want to point them out and explain our reasoning.

If you feel an exception should be made for some particular change, please put it on the agenda for the Glance weekly meeting so it can be discussed.

Database migration scripts

Once a database script has been included in a release, spelling or grammar corrections in comments are forbidden unless you are fixing them as a part of another stronger bug on the migration script itself. Modifying migration scripts confuses operators and administrators – we only want them to notice serious problems. Their preference must take precedence over fixing spell errors.

Typographical errors in comments

Comments are not user-facing. Correcting minor misspellings or grammatical errors only muddies the history of that part of the code, making git blame arguably less useful. So such changes are likely to be rejected. (This prohibition, of course, does not apply to corrections of misleading or unclear comments, or for example, an incorrect reference to a standards document.)

Misspellings in code

Misspellings in function names are unlikely to be corrected for the “historical clarity” reasons outlined above for comments. Plus, if a function is named mispelled() and a later developer tries to call misspelled(), the latter will result in a NameError when it’s called, so the later developer will know to use the incorrectly spelled function name.

Misspellings in variable names are more problematic, because if you have a variable named mispelled and a later developer puts up a patch where an updated value is assigned to misspelled, Python won’t complain. The “real” variable won’t be updated, and the patch won’t have its intended effect. Whether such a change is allowed will depend upon the age of the code, how widely used the variable is, whether it’s spelled correctly in other functions, what the current test coverage is like, and so on. We tend to be very conservative about making changes that could cause regressions. So whether a patch that corrects the spelling of a variable name is accepted is a judgment (or is that “judgement”?) call by reviewers. In proposing your patch, however, be aware that your reviewers will have these concerns in mind.

Tests

Occasionally someone proposes a patch that converts instances of assertEqual(True, whatever) to assertTrue(whatever), or instances of assertEqual(False, w) to assertFalse(w) in tests. Note that these are not type safe changes and they weaken the tests. (See the Python unittest docs for details.) We tend to be very conservative about our tests and don’t like weakening changes.

We’re not saying that such changes can never be made, we’re just saying that each change must be accompanied by an explanation of why the weaker test is adequate for what’s being tested.

Just to make this a bit clearer it can be shown using the following example, comment out the lines in the runTest method alternatively:

import unittest

class MyTestCase(unittest.TestCase):
    def setUp(self):
        pass

class Tests(MyTestCase):
    def runTest(self):
        self.assertTrue('True')
        self.assertTrue(True)
        self.assertEqual(True, 'True')

To run this use:

python -m testtools.run test.py

Also mentioned within the unittests documentation.

LOG.warn to LOG.warning

Consistently there are proposed changes that will change all {LOG,logging}. warn to {LOG,logging}.warning across the codebase due to the deprecation in Python 3. While the deprecation is real, Glance uses oslo_log that provides alias warn and solves the issue in single place for all projects using it. These changes are not accepted due to the huge amount of refactoring they cause for no reason.

Gratuitous use of oslo libraries

We are big fans of the oslo libraries and all the hard work the Oslo team does to keep common code reusable and easily consumable. But that doesn’t mean that it’s a bug if Glance isn’t using an oslo library everywhere you could possibly use one. We are all for using oslo if it provides any level of benefit for us and makes sense, but please let’s not have these bugs/patches of “Let’s use oslo because it exists”.