Going with option 2 is probably the best option so that you can use pytest fixtures and parameterization.
Might be worth looking at how Matplotlib re-arranged things on our master branch to maintain back-compatibility with nose-specific tools that were used by down-stream projects.TomOn Tue, Jul 11, 2017 at 4:22 PM Sebastian Berg <sebastian@sipsolutions.net> wrote:> https://mail.python.org/On Tue, 2017-07-11 at 14:49 -0600, Charles R Harris wrote:
> Hi All,
>
> Just looking for opinions and feedback on the need to keep NumPy from
> having a hard nose/pytest dependency. The options as I see them are:
>
> pytest is never imported until the tests are run -- current practice
> with nose
> pytest is never imported unless the testfiles are imported -- what I
> would like
> pytest is imported together when numpy is -- what we need to avoid.
> Currently the approach has been 1), but I think 2) makes more sense
> and allows more flexibility.
I am not quite sure about everything here. My guess is we can do
whatever we want when it comes to our own tests, and I don't mind just
switching everything to pytest (I for one am happy as long as I can run
`runtests.py` ;)).
When it comes to the utils we provide, those should keep working
without nose/pytest if they worked before without it I think.
My guess is that all your options do that, so I think we should take
the one that gives the nicest maintainable code :). Though can't say I
looked enough into it to really make a well educated decision, that
probably means your option 2.
- Sebastian
> Thoughts?
> Chuck
> _______________________________________________
> NumPy-Discussion mailing list
> NumPy-Discussion@python.orgmailman/listinfo/numpy- ____________________discussion ___________________________
NumPy-Discussion mailing list
NumPy-Discussion@python.org
https://mail.python.org/mailman/listinfo/numpy- discussion
_______________________________________________
NumPy-Discussion mailing list
NumPy-Discussion@python.org
https://mail.python.org/mailman/listinfo/numpy- discussion