Add back in explicit windows and mac testing

[paulf81]

This is probably only a draft pull request, I was curious to see if it could help shed light on on #951. Adding windows os (and mac) back into CI OS testing matrix

[paulf81]

Closing this one as seems like over-responding to one particular incident and greatly increasing our test count

[rafmudaf]

fwiw I think it's worth including Mac and Windows in the test matrix even if it's only for a subset. For example, you could run the example suite on all three systems but run the tests only on ubuntu. It would just be good to know that something is running on each OS.

[misi9170]

fwiw I think it's worth including Mac and Windows in the test matrix even if it's only for a subset. For example, you could run the example suite on all three systems but run the tests only on ubuntu. It would just be good to know that something is running on each OS.

This feels like a reasonable compromise (examples only on ubuntu, since those take a while, and tests on all three OSs, since those are faster). I'll reopen the PR and make those updates.

[misi9170]

After readding macOS and windows tests, it appears that the only failing tests are on Windows using python 3.9; and moreover, they are all related to using parallel operations (most are explicit, but note that LayoutOptimizationRandomSearch also defaults to parallelized operation). Evidently, everything works fine on python 3.10 and higher on Windows.

So, there are a couple of options here:

• Work on creating a fix so that the tests pass with python 3.9 and windows
• Drop support for python 3.9 (which will be at end of life later this year)

[rafmudaf]

Just a heads up @misi9170 that a description of the issue is at #951 (comment). Based on their response, it looks like like @RHammond2 agrees with requiring Python 3.10+. I could go either way on that.

In addition to the suggestions above, here are some more ways around this:

• Somehow require Python 3.10+ for Windows only
• Leave Issues when running FLORIS 4.1.1 #951 open with a "won't fix" tag so that other users will be able to understand the problem if it comes up (in this case, we'd skip or remove the test)

[paulf81]

I prefer dropping 3.9, it's kind of a pain for reasons beyond those in this issue. Sometimes tests crash just for 3.9 and it's because you need to add that from __future__ import annotations bit for 3.9 even though all other versions don't require it.

[RHammond2]

@rafmudaf, yes, I fully recommend moving to 3.10+. I've already moved a couple projects to 3.10+ at this point because 3.9 is out the door this year anyway, and it was causing some nuisance issues that were hard to resolve. Also, 3.10 introduces match statements, which can be a bit more robust than if/elif/else blocks: https://docs.python.org/3/tutorial/controlflow.html#match-statements and I haven't looked back.