New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Documentation builds fail - Mapgenerator breaks with matplotlib >= 3.6 #2800
Comments
eagle eyed Klaus 🦅 I was just about to open the same thing! Here is the test fail - currently tests pass since the environment has automagically reverted to matplotlib 3.5.3 (I have seen a bunch of unrelated failures in the 3.6.0 rc so prob the Matplotlib guys are looking into those atm); shall we pin Matplotlib for now? |
I suggest we find (even an interim) solution to this rather sooner than later, please - our tests are starting to drop like proverbial flies, see the conda lock test now - I can push a temp pin right away if that's OK with you, Manu is on holidays so maybe Saskia can give us some help with the BSC package, but in the meantime I suggest we pin the plotster |
I have reported that already. I will let you know when it's solved. |
cheers, Saskia 🍺 |
An example of a deprecation in ESMValCore can be found here: It basically uses the Sphinx directive An alternative could be the decorator |
Would it make sense to pin matplotlib for now as proposed by @valeriupredoi? Our documentation build appears to be failing because of this (see #2824). |
if there are no moves on this by Friday, I will open a PR with a matplotlib pin - but what's the situation with Mapgenerator at good old BSC, @sloosvel |
I don't think it will be ready for this release. You can follow or ask in this issue: https://earth.bsc.es/gitlab/es/mapgenerator/-/issues/80 |
thanks @sloosvel - the level of appetite for solving the issue seems to be somewhat minimal gauging from the discussion on that thread, I'll pin Matplotlib soon then 👍 |
good catch @bouweandela - why did silly GH API close this, I didn't say |
GitHub still needs to learn what 'do not close' means 😄 |
at least GH is not a grammar naz* 😆 |
@sloosvel, any progress on this upstream? I am keen to move forward with matplotlib. If I recall correctly, the fix is not terribly hard to do. Let me know if you need help. |
Let me ask again |
Ok version 1.0.6 is now available on pypi, the decorator that was causing issues is now removed. |
That's great! Would you or one of your colleagues like to become co-maintainer on the conda-forge feedstock? I'm happy to continue, so no pressure; just the bus factor. |
Mapgenerator is now released in a fixed version also on conda-forge, so we should be able to drop our pin on matplotlib. |
muzik to me ears, let me open a PR with the pin drop then - cheers for the good work, guys 🍻 |
FYI @fbeninca |
The
mapgenerator
package is incompatible with matplotlib >= 3.6 because it uses the functionmatplotlib.cbook.deprecated
, which has been deprecated in matplotlib 3.4 and has finally been removed in matplotlib 3.6; this breaks our tests.I will patch the conda-forge package of
mapgenerator
to to introduce an upper limit, but this is a temporary fix since it blocks us from updating matplotlib.For a permanent fix it would be good to fix mapgenerator itself.
@sloosvel, since the package originates from BSC, do you think you could drive this forward?
Perhaps @schlunma, you could help because, as I understand it, the problem is that it uses some API from matplotlib that was used to mark deprecations and that has been removed or made private to matplotlib as the general Python deprecation handling matured, and you have some experience with deprecations?
The text was updated successfully, but these errors were encountered: