On Tue, Jan 14, 2020 at 10:41 AM Timo Kaufmann <eisfre...@gmail.com> wrote: > > Am Montag, 13. Januar 2020 17:33:56 UTC+1 schrieb E. Madison Bray: >> >> On Sat, Jan 11, 2020 at 9:24 AM Antonio Rojas <nqn...@gmail.com> wrote: >> > >> > El viernes, 10 de enero de 2020, 14:54:24 (UTC+1), E. Madison Bray >> > escribió: >> >> >> >> That seems like the obvious approach to me. As it is I've long felt >> >> that Sage should be more flexible in its dependencies where >> >> possible/necessary. With most Python packages it's easy as most have >> >> a <package>.__version__ and its not so hard to define some variable >> >> like IS_RPY_2 and just have two separate branches. I have things like >> >> that all over the place in other packages to support e.g. different >> >> Numpy versions or work around version-specific bugs. >> > >> > >> > I've opened https://trac.sagemath.org/ticket/28988 for rpy. But at this >> > point the major issues are python 3.8 and ipython 7, and I don't see how >> > one could support several versions of them without forking hundreds of >> > doctests. Both updates require multi-thousand-lines patches, due to >> > changes in dict sorting and hashes. >> >> That remains a fault of over-reliance on doctests. > > > What else should we rely on? Also, doctests are not the only things that fail > with current python3 libraries.
Normal unit tests. The annoying thing about Sage's doctest-centric testing framework (which is otherwise good) is that it tends to lead people to thinking they can't write unit tests. But in fact there are unit tests in Sage, and they're just invoked through the doctest runner by writing "doctests" that run a unit test suite. >> I don't think >> downstream packaging is a good enough reason to push sage to rush >> things in such a way that is not well-communicated to the user >> community. If you need to have a multi-thousand-line patch then so be >> it. A patch is a patch. > > > That is unfortunate. I agree that "a patch is a patch", but in my view the > conclusion should be the opposite: Upstream should strive for no patches to > be necessary (except maybe to work around very distro-specific quirks). No > 5000 line patches and no 5 line patches. > > For me this decision means that sage on nixos will likely be stuck on python2 > for a while, and I can only hope that the python2 infrastructure keeps > working for long enough. Well, all the more reason to maintain Python 2 support a little longer than to rush and break things. -- You received this message because you are subscribed to the Google Groups "sage-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/CAOTD34YreBJwmQS97R5%2Bn2YmpDxeTR8OJ_42Y7G3nQZCktH6Mw%40mail.gmail.com.
