tests: remove private magicgui access in tests

[tlambert03]

Description

This removes a reference to a private module-level variable in magicgui that is going to be moved in pyapp-kit/magicgui#497. It's only used in the tests, and the specific test (resolution of forward references) is well tested on the magicgui side at this point. So simply removing it is fine

[alisterburt]

awesome, thanks for taking the time to clear up here!

[codecov]

Codecov Report

Merging #5331 (2d7475c) into main (cd5c314) will increase coverage by 0.00%.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##             main    #5331   +/-   ##
=======================================
  Coverage   89.08%   89.09%           
=======================================
  Files         582      582           
  Lines       49311    49310    -1     
=======================================
+ Hits        43929    43932    +3     
+ Misses       5382     5378    -4     

Impacted Files	Coverage Δ
napari/_tests/test_magicgui.py	96.42% <100.00%> (-0.02%)	⬇️
napari/utils/theme.py	89.79% <0.00%> (+0.68%)	⬆️
napari/utils/info.py	83.33% <0.00%> (+1.11%)	⬆️
napari/_qt/__init__.py	55.17% <0.00%> (+6.89%)	⬆️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[Czaki]

But if I understand correctly, the whole test should be removed based on its description:

object triggers the appropriate imports to resolve the class in time.

[tlambert03]

I can do that too if that's what you prefer

[tlambert03]

ok removed

[tlambert03]

I had to revert the removal since unrelated tests (test_expected_number_of_layer_methods ) started failing:

_______________ test_expected_number_of_layer_methods[Image-5] ________________
[305](https://github.com/napari/napari/actions/runs/3455538721/jobs/5767631644#step:8:306)
  
[306](https://github.com/napari/napari/actions/runs/3455538721/jobs/5767631644#step:8:307)
  cls = 'Image', expectation = 5
[307](https://github.com/napari/napari/actions/runs/3455538721/jobs/5767631644#step:8:308)
  
[308](https://github.com/napari/napari/actions/runs/3455538721/jobs/5767631644#step:8:309)
      @pytest.mark.parametrize(
[309](https://github.com/napari/napari/actions/runs/3455538721/jobs/5767631644#step:8:310)
          'cls, expectation', EXPECTED_NUMBER_OF_LAYER_METHODS.items()
[310](https://github.com/napari/napari/actions/runs/3455538721/jobs/5767631644#step:8:311)
      )
[311](https://github.com/napari/napari/actions/runs/3455538721/jobs/5767631644#step:8:312)
      def test_expected_number_of_layer_methods(cls, expectation):
[312](https://github.com/napari/napari/actions/runs/3455538721/jobs/5767631644#step:8:313)
          """
[313](https://github.com/napari/napari/actions/runs/3455538721/jobs/5767631644#step:8:314)
          Make sure we do find all the methods attached to a layer via keybindings
[314](https://github.com/napari/napari/actions/runs/3455538721/jobs/5767631644#step:8:315)
          """
[315](https://github.com/napari/napari/actions/runs/3455538721/jobs/5767631644#step:8:316)
          layer_methods = _get_all_keybinding_methods(getattr(layers, cls))
[316](https://github.com/napari/napari/actions/runs/3455538721/jobs/5767631644#step:8:317)
  >       assert len(layer_methods) == expectation
[317](https://github.com/napari/napari/actions/runs/3455538721/jobs/5767631644#step:8:318)
  E       AssertionError: assert 8 == 5
[318](https://github.com/napari/napari/actions/runs/3455538721/jobs/5767631644#step:8:319)
  E        +  where 8 = len({'activate_image_pan_zoom_mode', 'activate_image_select_mode', 'hold_to_pan_zoom', 'orient_plane_normal_along_view_direction', 'orient_plane_normal_along_x', 'orient_plane_normal_along_y', ...})

I have never quite understood where those seemingly magic expectation numbers come from... it appears that in general/historicall we just change them to make tests pass... which kinda defeats the purpose of the test.

It's out of scope for this PR. Sorry @Czaki.

[Czaki]

It's out of scope for this PR. Sorry @Czaki.

Ah. There is a test that uses the side effect of another test to pass... Maybe we should have a test turn that runs each test in an independent python interpreter?

In the context of this PR I'm fine, but it would be nice if you could update the docstring of test function to reflect that it is for validate if the expected magicgui widget is returned.

[tlambert03]

There is a test that uses the side effect of another test to pass... Maybe we should have a test turn that runs each test in an independent python interpreter?

exactly, and it's just generally been updated to make it pass given whatever the current side effects are 😂 ... that whole test should probably be removed/refactored. it doesn't appear to be a complicated thing to check, but it needs to be redesigned (in another PR)