remove nose-style test classes #24148

Activity

1. tacaswell

   added

   Difficulty: Mediumhttps://matplotlib.org/devdocs/devel/contribute.html#good-first-issueshttps://matplotlib.org/devdocs/devel/contribute.html#good-first-issues

   🌱 Good first issueOpen a pull request against these issues if there are no active ones!Open a pull request against these issues if there are no active ones!

   on Oct 12, 2022
2. tacaswell

   added this to the v3.7.0 milestone on Oct 12, 2022
3. elrorvik commented on Oct 13, 2022

   

   elrorvik

   on Oct 13, 2022

   Last edited by elrorvik

   More actions

   Hi! I'm trying to find the nose-settings in the pytest-config, but I could not find anything about nose-library except in the docs...
   
   Could you point me in the direction of the nose-configs? (From what I could find it was supposed to be in this contest: https://github.com/matplotlib/matplotlib/blob/main/lib/matplotlib/testing/conftest.py)
4. tacaswell commented on Oct 13, 2022

   

   tacaswell

   on Oct 13, 2022

   MemberAuthor

   More actions

   I think the issue is that we are implicitly using nose-style classes that pytest has in the past happily discovered and ran them, but on the current development branch of pytest is starting to warn that this will be removed in the future.

   We have our tests set to fail on warning so this is causing test failures for us now (or well as soon as the next version of pytest is released). Hopefully we can get this addressed before then (by fixing it on main and pinning on the bug-fix branch) so we never see these failures on our CI and do not have to deal with this when it becomes a crisis.
5. oscargus commented on Oct 14, 2022

   

   oscargus

   on Oct 14, 2022

   Last edited by oscargus

   Member

   More actions

   I think that only test classes with a setup method count as a Nose test?

   For e.g. the Test_boxplot_stats class
   

   matplotlib/lib/matplotlib/tests/test_cbook.py

   Lines 53 to 176 in f0cd8de

   	class Test_boxplot_stats:
   	def setup(self):
   	np.random.seed(937)
   	self.nrows = 37
   	self.ncols = 4
   	self.data = np.random.lognormal(size=(self.nrows, self.ncols),
   	mean=1.5, sigma=1.75)
   	self.known_keys = sorted([
   	'mean', 'med', 'q1', 'q3', 'iqr',
   	'cilo', 'cihi', 'whislo', 'whishi',
   	'fliers', 'label'
   	])
   	self.std_results = cbook.boxplot_stats(self.data)
   	self.known_nonbootstrapped_res = {
   	'cihi': 6.8161283264444847,
   	'cilo': -0.1489815330368689,
   	'iqr': 13.492709959447094,
   	'mean': 13.00447442387868,
   	'med': 3.3335733967038079,
   	'fliers': np.array([
   	92.55467075, 87.03819018, 42.23204914, 39.29390996
   	]),
   	'q1': 1.3597529879465153,
   	'q3': 14.85246294739361,
   	'whishi': 27.899688243699629,
   	'whislo': 0.042143774965502923
   	}
   	self.known_bootstrapped_ci = {
   	'cihi': 8.939577523357828,
   	'cilo': 1.8692703958676578,
   	}
   	self.known_whis3_res = {
   	'whishi': 42.232049135969874,
   	'whislo': 0.042143774965502923,
   	'fliers': np.array([92.55467075, 87.03819018]),
   	}
   	self.known_res_percentiles = {
   	'whislo': 0.1933685896907924,
   	'whishi': 42.232049135969874
   	}
   	self.known_res_range = {
   	'whislo': 0.042143774965502923,
   	'whishi': 92.554670752188699
   	}
   	def test_form_main_list(self):
   	assert isinstance(self.std_results, list)
   	def test_form_each_dict(self):
   	for res in self.std_results:
   	assert isinstance(res, dict)
   	def test_form_dict_keys(self):
   	for res in self.std_results:
   	assert set(res) <= set(self.known_keys)
   	def test_results_baseline(self):
   	res = self.std_results[0]
   	for key, value in self.known_nonbootstrapped_res.items():
   	assert_array_almost_equal(res[key], value)
   	def test_results_bootstrapped(self):
   	results = cbook.boxplot_stats(self.data, bootstrap=10000)
   	res = results[0]
   	for key, value in self.known_bootstrapped_ci.items():
   	assert_approx_equal(res[key], value)
   	def test_results_whiskers_float(self):
   	results = cbook.boxplot_stats(self.data, whis=3)
   	res = results[0]
   	for key, value in self.known_whis3_res.items():
   	assert_array_almost_equal(res[key], value)
   	def test_results_whiskers_range(self):
   	results = cbook.boxplot_stats(self.data, whis=[0, 100])
   	res = results[0]
   	for key, value in self.known_res_range.items():
   	assert_array_almost_equal(res[key], value)
   	def test_results_whiskers_percentiles(self):
   	results = cbook.boxplot_stats(self.data, whis=[5, 95])
   	res = results[0]
   	for key, value in self.known_res_percentiles.items():
   	assert_array_almost_equal(res[key], value)
   	def test_results_withlabels(self):
   	labels = ['Test1', 2, 'Aardvark', 4]
   	results = cbook.boxplot_stats(self.data, labels=labels)
   	for lab, res in zip(labels, results):
   	assert res['label'] == lab
   	results = cbook.boxplot_stats(self.data)
   	for res in results:
   	assert 'label' not in res
   	def test_label_error(self):
   	labels = [1, 2]
   	with pytest.raises(ValueError):
   	cbook.boxplot_stats(self.data, labels=labels)
   	def test_bad_dims(self):
   	data = np.random.normal(size=(34, 34, 34))
   	with pytest.raises(ValueError):
   	cbook.boxplot_stats(data)
   	def test_boxplot_stats_autorange_false(self):
   	x = np.zeros(shape=140)
   	x = np.hstack([-25, x, 25])
   	bstats_false = cbook.boxplot_stats(x, autorange=False)
   	bstats_true = cbook.boxplot_stats(x, autorange=True)
   	assert bstats_false[0]['whislo'] == 0
   	assert bstats_false[0]['whishi'] == 0
   	assert_array_almost_equal(bstats_false[0]['fliers'], [-25, 25])
   	assert bstats_true[0]['whislo'] == -25
   	assert bstats_true[0]['whishi'] == 25
   	assert_array_almost_equal(bstats_true[0]['fliers'], [])

   
   one can either just create the data in every method (now a function) or, the more advanced way, create a pytest fixture with the data.

   For Test_callback_registry it seems to make sense to create a fixture.

   For TestDetrend either create data or fixture. (Some of the tests seems viable for parameterization as well).

   I guess the thing to closely look for is coverage changes. If there are any, something went wrong. If the coverage it still the same, there is a chance it is correct.
6. rcomer commented on Oct 14, 2022

   

   rcomer

   on Oct 14, 2022

   Member

   More actions

   Would it be enough to just replace setup with setup_method?
   https://docs.pytest.org/en/7.1.x/how-to/xunit_setup.html?highlight=setup_method#method-and-function-level-setup-teardown

   👍React with 👍6Reacted by hannah, Tim Hoffmann, Oscar Gustafsson, Armando, Pankaj Singh and Sudheer Reddy
7. timhoffm

   closed this as completedin #24173on Oct 15, 2022
8. QuLogic

   modified the milestones: v3.7.0, v3.6.2 on Oct 17, 2022