address shapely 2.0 deprecations

[akrherz]

Rationale

Addresses ShapelyDeprecationWarnings raised with the newly released Shapely 1.8 that is preparing for the upcoming Shapely 2.0 breaking API release.

Implications

No known implications nor previously supported shapely version conflicts.

fixes #1913

NB: My fork CI still emits three warnings, I am unsure if I am missing things yet or this is in other packages

Update, I ran pytest once with -W error::shapely.errors.ShapelyDeprecationWarning and believe I have found the source of the three previously listed warnings below.

• /usr/share/miniconda/envs/test-environment/lib/python3.8/site-packages/numpy/testing/_private/utils.py:1046: ShapelyDeprecationWarning: The array interface is deprecated and will no longer work in Shapely 2.0. Convert the '.coords' to a numpy array instead.
• <array_function internals>:5: ShapelyDeprecationWarning: The array interface is deprecated and will no longer work in Shapely 2.0. Convert the '.coords' to a numpy array instead.
• /usr/share/miniconda/envs/test-environment/lib/python3.8/site-packages/numpy/lib/function_base.py:4817: ShapelyDeprecationWarning: The array interface is deprecated and will no longer work in Shapely 2.0. Convert the '.coords' to a numpy array instead.

[akrherz]

Darn, CI error py37+Mac appears to be unrelated to this PR.

[akrherz]

I'm gonna track down those three warnings yet shown and then mark the PR ready once done.

[akrherz]

OK and whew, I think I have found all the ShapelyDeprecationWarnings now.

[QuLogic]

Hmm, it looks like the Minimum dependency CI run does not pin Shapely. We should probably do that to be sure that this doesn't break older versions.

[akrherz]

We should probably do that to be sure that this doesn't break older versions.

@QuLogic I think I have accomplished that now

[QuLogic]

Thanks @akrherz!

[jorisvandenbossche]

Thanks for already addressing those warnings! (and glad that it didn't seem too problematic to address them)

A question and suggestion:

• Would this be considered for a 0.20.2 release? (I don't know how far a 0.21 release is) Because right now everybody installing the latest versions of packages will see lots of warnings
• I would suggest to add a upper pin to the shapely dependency in the install_requirements to be <2. That would prevent users to run into problems in the future if Shapely 2.0 is released and if the released version of cartopy at that moment does not yet support it (which is currently not yet the case, since you are also using shapely.geos.lgeos, for which we didn't add a deprecation warning ... updating for this is related to Minimizing incompatibility between cartopy and shapely #805 (comment))

[greglucas]

Seems reasonable to include this in a patch release. I also don't have a huge issue with making a new minor release either.

[greglucas]

@akrherz, do you want to point these two commits at v0.20.x? We can also add the shapely pin to that branch too then.

[akrherz]

@akrherz, do you want to point these two commits at v0.20.x?

Hopefully I did what you requested in #1930