MNT: bump python versions and dependency minimum versions #2169
Conversation
That seems likely (I'm by no means an expert on this!) I updated the cython version here to match the min version tests. The setuptools version here is from 2018 and setuptools_scm only last year, so not sure they were previously following the ~2 year policy. I removed the mpl version check from that test, and also found some handling for older mpl versions in the main code. |
| # The shading keyword argument was added in MPL 3.3, so keep | ||
| # this default updating until we only support MPL>=3.3 | ||
| default_shading = mpl.rcParams.get('pcolor.shading', 'auto') | ||
| default_shading = mpl.rcParams.get('pcolor.shading') |
There was a problem hiding this comment.
I'm not sure how to read the above comment... Can we get rid of the default altogether now and assume kwargs contains "shading" now?
| default_shading = mpl.rcParams.get('pcolor.shading') |
There was a problem hiding this comment.
I think the user can just not pass shading, or am I misunderstanding the question?
There was a problem hiding this comment.
Oh yeah, I was confused thinking there was a default kwarg pcolormesh(*, shading="auto") that we could potentially rely on. But, I think you're right we need to keep this because we are intercepting the user kwargs before going to the super() call.
| @@ -44,7 +45,7 @@ | |||
| from cartopy.mpl.slippy_image_artist import SlippyImageArtist | |||
|
|
|||
|
|
|||
| assert mpl.__version__ >= '3.4', \ | |||
| assert packaging.version.parse(mpl.__version__).release[:2] >= (3, 4), \ | |||
There was a problem hiding this comment.
I guess mpl v3.10 will ship next year, at which point the string comparison would become problematic.
Rationale
Python 3.8 recently fell off the NEP29 schedule, so this PR removes testing with that and also adds it for python 3.11.
https://numpy.org/neps/nep-0029-deprecation_policy.html
I also bumped versions of other packages used in the minimum versions test as I assume bad things can happen if you bump the python version without doing this! I wasn't sure exactly which versions to use here but mostly went for the oldest that is newer than 2 years. For shapely it seemed safer to go with the older 1.7.1 (August 2020) as 1.8.x had many patch releases stretching into last year. Matplotlib 3.4 is also older than 2 years but 3.5 didn't come out till November 2021.
Implications