Conversation
|
... if we make anything a new core dep in the future, we could consider doing it with nibabel. We could remove a lot of checks by doing it. And maintenance overhead shouldn't be bad because it's a pure-python dependency with a slow-moving deprecation cycle. |
drammock
left a comment
There was a problem hiding this comment.
git grep matplotlib.*version shows a few more:
mne/fixes.py:923: if _compare_version(matplotlib.__version__, '>=', '3.2.0'):
mne/viz/_brain/_scraper.py:74: _compare_version(matplotlib.__version__, '>=', '3.3.1') and
mne/viz/_mpl_figure.py:1324: # XXX when matplotlib 3.3 is min version, replace this with
mne/viz/tests/test_topomap.py:244: # ...but not all matplotlib versions support it, and we can't use
I can tackle those if you don't have time. |
|
Nice catch! I think I got those now |
I'll revert that last commit. The comment didn't specify/know the matplotlib version where we could use the simpler version... apparently it's something after 3.4! |
|
pydata-sphinx-theme building failed on Windows and 0.13.1 is now out (as of about a week ago) so I'm going to pin to that. It gives us about a month to see issues with dev docs and either fix them before release or relax/revert the pin. In the meantime it'll speed up the doc builds |
FYI I've been (quietly) trying to fix our site config to look good with 0.13.1. The one thing I haven't been able to fix yet is that we somehow lost margin/padding between the gallery tiles on the homepage, but it actually happened a while ago and I haven't had time to track down what PST change triggered it (I suspect it was the bootstrap4 -> bootstrap5 migration). So feel free to ignore that particular doc issue. |
|
Circle is having a bad day so merged manually |
* upstream/main: MAINT: Bump minimums (mne-tools#11557) Bump minimum Python to 3.8 (mne-tools#11550)
Bumps min reqs in README.rst to those ~2 years old this April, and adjusts/removes code shims for deps lower than those. Also
nibabelis now required forread_surfacewhich I think is reasonable (it is pure Python and has been stable and pip/conda-installable for years now).Closes #10977