[SciPy-Dev] Issues backlog

[Ralf Gommers]

On Mon, Aug 5, 2019 at 8:36 PM Eric Larson <larson.eric.d at gmail.com> wrote:

> Thanks for working through those issues (and more at this very moment it
> appears!).
>

yeah, it feels kind of productive:) down from 1242 issues 2 days ago to
1185 issues right now. and I believe there's enough useful info in some of
the older issues that it's not easily possible to close them automatically.

I'll try to make time for this sort of thing as well.
>

thanks!

Ralf



> Initially I was overall in favor of the auto-closing-after-a-warning bot,
> but the continued discussion here has persuaded me that manual intervention
> might actually be better (higher SNR and more efficient) in the end.
>
> Eric
>
>
> On Mon, Aug 5, 2019 at 10:28 PM Ralf Gommers <ralf.gommers at gmail.com>
> wrote:
>
>>
>>
>> On Mon, Aug 5, 2019 at 10:39 AM Ralf Gommers <ralf.gommers at gmail.com>
>> wrote:
>>
>>>
>>>
>>> On Sun, Aug 4, 2019 at 1:56 PM Stefan van der Walt <stefanv at berkeley.edu>
>>> wrote:
>>>
>>>> On Sat, Aug 3, 2019, at 22:22, Dieter Werthmüller wrote:
>>>> > I think this is a very good idea, and it should be applied to issues
>>>> and
>>>> > PRs. However, instead of age as a criteria I would use inactivity
>>>> (e.g.,
>>>> > if there is an old issue from 2013 that has activity discussions every
>>>> > year then it should not be closed).
>>>>
>>>> Agreed, the last modified age may be a better measure.  That way we
>>>> don't need to recreate issues we want to keep open, we can simply leave a
>>>> comment / add a label / edit the description.
>>>>
>>>> But this method for closing issues can also be infuriating to users:
>>>> how many times have you come across a project where an issue was described
>>>> in detail with debugging info, only to be closed by a bot due to
>>>> inactivity?  Perhaps that can be addressed by setting the
>>>> inactivity-time-to-closure to a high enough duration, perhaps 3 years.
>>>>
>>>
>>> I agree with the infuriating part: each time I've had an experience like
>>> that with another project (e.g. pip), it has been extremely annoying. To
>>> the extent I decided to never contribute again. Issues closed that were
>>> valid, PRs closed that didn't get reviewed, etc. It is a good way to tell
>>> contributors that their contributions aren't valued, and/or lose useful
>>> information.
>>>
>>> In general, valid bug reports should simply not be closed imho. The
>>> exception is probably new feature requests. About 300 of the 1200 open
>>> issues have the "enhancement" label. Looking through the older enhancements
>>> shows that many can be closed. However even there, there's useful content
>>> in some. I'd much rather go through them and close by hand than have some
>>> bot do it. This has multiple advantages:
>>> - doesn't anger contributors
>>> - keeps the useful ones
>>> - is probably _less_ work (at 3 minutes per issue one could triage the
>>> 200 issues that haven't been touched in the last 2 years in a single day),
>>> choosing and maintaining a bot is easily going to take someone that much
>>> time.
>>>
>>
>> To validate this, I just reserved 30 minutes to go through enhancement
>> issues, starting from the oldest ones. Results: closed 10 (with
>> rationales), kept 1 open, pinged the relevant person on 2 to check whether
>> the issue could be closed. Full list:
>>
>> gh-788: 2 min
>> gh-818: 3 min
>> gh-857: 3 min relevant, leaving open
>> gh-881: 2 min
>> gh-885: 2 min
>> gh-983: 2 min, added question on leaving open or not
>> gh-1002: 3 min
>> gh-1005: 2 min
>> gh-1089: 1 min
>> gh-1175: 2 min, added question on leaving open or not
>> gh-1219: 2 min
>> gh-1335: 1 min
>> gh-1338: 3 min
>>
>> So my 3 minutes per issue was a conservative estimate. Newer issues could
>> require more time, but on the other hand I just looked at all issues in
>> order rather than picking the ones only from modules I'm more familiar
>> with. YMMV, but imho we can do a major cleanup of the issue list fairly
>> easily.
>>
>> Cheers,
>> Ralf
>>
>>
>>
>>
>>>
>>>
>>>> If this method is enacted, I suggest an email to the list once a month
>>>> with issues that were automatically closed.  That may at least provide an
>>>> opportunity for those interested to go back and "save" any important ones
>>>> they care about.
>>>>
>>>
>>> This is actually even more work. It asks all maintainers and mailing
>>> list contributors to triage with such an email, so many people will be
>>> doing duplicate work.
>>>
>>> Cheers,
>>> Ralf
>>>
>>> _______________________________________________
>> SciPy-Dev mailing list
>> SciPy-Dev at python.org
>> https://mail.python.org/mailman/listinfo/scipy-dev
>>
> _______________________________________________
> SciPy-Dev mailing list
> SciPy-Dev at python.org
> https://mail.python.org/mailman/listinfo/scipy-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/scipy-dev/attachments/20190805/c1805986/attachment-0001.html>