Revive pylint support, and a first batch for fixes for problems found

[ydirson]

pylint will be of help for python3 support.

[edwintorok]

Thanks for enabling pylint. Could we get it to run using github actions?
See here what SM does: https://github.com/xapi-project/sm/blob/master/.github/workflows/main.yml

Might be useful to have a test matrix there: test using both python2.7 and 3.x to ensure our changes are compatible with both.

[edwintorok]

I think this can be done via some of the @abc annotations in python, would https://docs.python.org/3/library/abc.html#abc.abstractmethod (and inheriting from the proper ABC class) fix the pylint warning too?

[ydirson]

Yes, using abc works just as well. That one is annoying in projects making heavy use of metaclasses (which is why I default to the exception idiom), but it will do the job here.

[ydirson]

Note, inheriting from abc.ABC comes with 3.4 only, so we'd have to use something like @six.add_metaclass(abc.ABCMeta) for a 2/3-portable idiom, which I feel brings too much complexity, for no clear advantage over NotImplementedError(). What do you think ?

[MarkSymsCtx]

Note, inheriting from abc.ABC comes with 3.4 only, so we'd have to use something like @six.add_metaclass(abc.ABCMeta) for a 2/3-portable idiom, which I feel brings too much complexity, for no clear advantage over NotImplementedError(). What do you think ?

I think I'd concur with that. My feeling is we want the Py2/Py3 compatibility phase to be relatively short until the migration is complete and then drop the py2 support completely at which point there can be a follow on set of changes to perform code cleanups.

[ydirson]

Thanks for enabling pylint. Could we get it to run using github actions? See here what SM does: https://github.com/xapi-project/sm/blob/master/.github/workflows/main.yml

sure!

Might be useful to have a test matrix there: test using both python2.7 and 3.x to ensure our changes are compatible with both.

That does make sense for unit tests, but for pylint the usefulness will I think mostly be in making use of newer py3-only versions of pylint - all of which requires to get the migration far enough first.

[edwintorok]

I think to enable the CI we'd have to merge this to master, I see it passed on your fork

[MarkSymsCtx]

Note, inheriting from abc.ABC comes with 3.4 only, so we'd have to use something like @six.add_metaclass(abc.ABCMeta) for a 2/3-portable idiom, which I feel brings too much complexity, for no clear advantage over NotImplementedError(). What do you think ?

I think I'd concur with that. My feeling is we want the Py2/Py3 compatibility phase to be relatively short until the migration is complete and then drop the py2 support completely at which point there can be a follow on set of changes to perform code cleanups.