addopting pytest-skip-slow
16 views
Skip to first unread message
Clément Robert
May 27, 2026, 4:05:48 PM (5 days ago) May 27
to astro...@googlegroups.com
Hi team
pytest-skip-slow, authored by Brian Okken, is now compatible with pytest-astropy's pytest plugin, without the extra bag of (dead) dependencies.
I propose we adopt the package as a replacement within astropy's own dependency tree.
full context: https://github.com/astropy/astropy/issues/18783I do note that pytest-skip-slow is currently a one-man show without an org attached, so I asked Brian if he'd like us to take over, moving the repo to the astropy organization, which he agreed to, on the condition that he can still share maintainership (which of course seems perfectly reasonable to me). What do you guys think ?
Clément
Marten van Kerkwijk
May 28, 2026, 5:13:49 AM (5 days ago) May 28
to astro...@googlegroups.com
Hi Clément,
I'm in favour, and had a quick look at pytest-skip-slow -- it seems
essentially a simply file with definitions, and I like the idea of
having that in the astropy organization, replacing the code in
pytest-astropy.
More generally, I agree with the move to explicit test dependencies,
where specific packages just do one thing well, not *also* pull in a
whole slew of other stuff.
It is perhaps worth thinking what the purpose of pytest-astropy was, and
whether that still serves a need. I think it has evolved to become a
metapackage that allows one to set up testing for a package in the way
its done for astropy. I've certainly used it myself for that purpose,
but for my more mature packages, I've gone to more limited dependencies
that I actually use (like pytest-doctestplus and
pytest-filter-subpackage). So, for me at least, pytest-astropy pulls in
more than is needed.
On the other hand, pytest-astropy does not pull in everything astropy
testing needs, so is incomplete in that sense. To make it more useful,
one could perhaps add more dependencies to it, but, really, that use
case is more easily served by using `astropy[test]` as a test
dependendency (that assumes one's package uses astropy, but, if not,
then it would seem outside the astropy project's scope).
Overall, to me it seems pytest-astropy no longer has a compelling use
case, and (slowly) moving towards deprecating it seems sensible.
Replacing the code it has with pytest-skip-slow seems a good step also
from that perspective.
All the best,
Marten
I'm in favour, and had a quick look at pytest-skip-slow -- it seems
essentially a simply file with definitions, and I like the idea of
having that in the astropy organization, replacing the code in
pytest-astropy.
More generally, I agree with the move to explicit test dependencies,
where specific packages just do one thing well, not *also* pull in a
whole slew of other stuff.
It is perhaps worth thinking what the purpose of pytest-astropy was, and
whether that still serves a need. I think it has evolved to become a
metapackage that allows one to set up testing for a package in the way
its done for astropy. I've certainly used it myself for that purpose,
but for my more mature packages, I've gone to more limited dependencies
that I actually use (like pytest-doctestplus and
pytest-filter-subpackage). So, for me at least, pytest-astropy pulls in
more than is needed.
On the other hand, pytest-astropy does not pull in everything astropy
testing needs, so is incomplete in that sense. To make it more useful,
one could perhaps add more dependencies to it, but, really, that use
case is more easily served by using `astropy[test]` as a test
dependendency (that assumes one's package uses astropy, but, if not,
then it would seem outside the astropy project's scope).
Overall, to me it seems pytest-astropy no longer has a compelling use
case, and (slowly) moving towards deprecating it seems sensible.
Replacing the code it has with pytest-skip-slow seems a good step also
from that perspective.
All the best,
Marten
Stuart Mumford
Jun 1, 2026, 10:11:30 AM (16 hours ago) Jun 1
to astro...@googlegroups.com
As always +1 to deleting code.
--
You received this message because you are subscribed to the Google Groups "astropy-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to astropy-dev...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/astropy-dev/CACMcR71pE3ZRA%3DJNXMuCccXpse5aVWVfQY9HXi0r3idEdQ%2Bpog%40mail.gmail.com.
Reply all
Reply to author
Forward
0 new messages