Add package Cartopy

[juntyr]

Description

Cartopy is a commonly used library to draw maps with matplotlib. Since most of its dependencies are already in pyodide, except for the pure Python dependency pyshp, adding cartopy turned out to be quite simple.

Cartopy often downloads shape files the first time that the user requires them during plotting, which is only supported if the pyodide-http patches are applied. How should this be handled?

Checklists

• Add a CHANGELOG entry
• Add / update tests
• Add new / update outdated documentation

[juntyr]

Mmmm, I'm not sure if I understand why the tests fail - could someone perhaps give me a pointer in the right direction? Thanks :)

[juntyr]

@hoodmane Ok, I've made some progress and may need your help again with decyphering and fixing the remaining errors. Can matplotlib be used in a test or are there some restrictions I should be mindful of?

[ryanking13]

Thanks @juntyr!

Cartopy often downloads shape files the first time that the user requires them during plotting, which is only supported if the pyodide-http patches are applied. How should this be handled?

I think using pyodide-http makes sense.

Can matplotlib be used in a test or are there some restrictions I should be mindful of?

You can see matplotlib tests for a reference when writing tests that use matplotlib.
You'll need to add some tests that check if the images (plots) are correctly created in Pyodide.

For example, you can run the code that generates an image (plot) locally, save the image into a file, then compare it with the same image created from Pyodide. You can copy the helper functions from matplotlib tests.

conftest.py:202: in pytest_runtest_call
    yield from extra_checks_test_wrapper(browser, trace_hiwire_refs, trace_pyproxies)
conftest.py:233: in extra_checks_test_wrapper
    assert (delta_proxies, delta_keys) == (0, 0) or delta_keys < 0
E   assert ((16, 34) == (0, 0)

You can skip this pyproxy check by adding the following decorators (again, check matplotlib tests to see how these decorators are applied). The error means there are some leaked Python objects after the test and it is quite normal when testing a package that interacts with JS.

pytest.mark.skip_refcount_check
pytest.mark.skip_pyproxy_check

[ryanking13]

Thanks. @juntyr! Please update the changelog.

[ryanking13]

Thanks again @juntyr!

[juntyr]

Thanks for the help and reviews @ryanking13 and @hoodmane!