Use six.integer_types everywhere

[Marc Tamlyn]

There are 23 places in the source code where we use isinstance(..., int) rather than six.integer_types. There is probably no reason why these should not accept instances of long on python 2, so we should update these instances (with regression tests) to be six.integer_types.

[bowlofkashi]

Hello,

This will be my first contribution to Django! I have found the occurrences of isinstance(..., int) and have changed them to isinstance(..., six.integer_types). I ran tests on Django using the command PYTHONPATH=..:$PYTHONPATH ./runtests.py

(I could not figure out how to make the bolded text monospaced, so I made it bold. Sorry!)

[bowlofkashi]

Sorry. That was a bad patch. I am making improvements. I could not find how to remove the bad patch file attachment. Please don't use it.

[bowlofkashi]

Patch for #22840

[bowlofkashi]

django-tests log file

[Tim Graham]

Thanks for your work so far. As noted in the description, we need regression tests for these modifications. That is for each change, there should be a test that fails before the change and passes afterward. That said, I am not convinced we should make these changes everywhere especially in places like logging where I don't think the logging level would every be a long(). Further, we'd have to skip the tests on Python 3 since long() doesn't exist there. Once we drop Python 2, we'd then just revert this change since it's no longer needed. All in all, it seems like a lot of perhaps unnecessary work to blindly make this change without some justification for each case (e.g. #22820).

As a new contributor, you may interested in our ​patch review checklist. By the way, there's no need to attach a log of the test output.

[bowlofkashi]

Replying to timo:
Thank you for your response and your help. I will read the patch review checklist, as well as the guidelines set for the project.

[Baptiste Mispelon]

Work has started on a branch here: ​https://github.com/riccardotonini/django/tree/ticket-22840

[Tim Graham]

Is there any rebuttal to my comment 6 where I questioned the usefulness of this?

[Aymeric Augustin]

Like timo, I'm skeptical about this change.

If someone wants to audit each case and see if longs could cause an issue, why not.

But if it was a common cause for bugs we would have heard about it before.