def test_sample_neighbors_locale_w_seed(self):
exp_metadata = self.context_md1
# since we're setting a random seed, the result we get the first
# time is our expected every time
exp_sel = sample_neighbors(self.focal_seqs1,
self.context_seqs1,
percent_id=0.98,
samples_per_cluster=2,
locale=self.context_md1.get_column('x'),
seed=0)
self.assertTrue(exp_sel.inclusion['c1'])
self.assertEqual(exp_sel.inclusion.sum(), 3)
self.assertEqual(exp_sel.metadata, exp_metadata)
for _ in range(self._N_TEST_ITERATIONS):
sel = sample_neighbors(self.focal_seqs1,
self.context_seqs1,
percent_id=0.98,
samples_per_cluster=2,
locale=self.context_md1.get_column('x'),
seed=0)
pdt.assert_series_equal(sel.inclusion,
exp_sel.inclusion)
def test_sample_neighbors_metadata_subset(self):
context_md = self.get_data_path('context-metadata-1-missing-id.tsv')
context_md = qiime2.Metadata.load(context_md)
with self.assertRaisesRegex(ValueError, 'not present in the metadata'):
sample_neighbors(self.focal_seqs1,
self.context_seqs1,
percent_id=0.98,
samples_per_cluster=1,
locale=context_md.get_column('x'))
def test_sample_neighbors_terminal_gaps_ignored(self):
sel = sample_neighbors(self.focal_seqs2,
self.context_seqs2,
percent_id=1.0,
samples_per_cluster=2)
exp_inclusion = pd.Series([True], index=['c1'], name='inclusion')
exp_metadata = pd.DataFrame(index=['c1'])
exp_metadata.index.name = 'id'
exp_metadata = qiime2.Metadata(exp_metadata)
pdt.assert_series_equal(sel.inclusion, exp_inclusion)
self.assertEqual(sel.metadata, exp_metadata)
self.assertEqual(sel.label, 'sample_neighbors')
def test_sample_neighbors_no_locale_alt_samples_per_cluster(self):
sel = sample_neighbors(self.focal_seqs1,
self.context_seqs1,
percent_id=0.98,
samples_per_cluster=3)
exp_inclusion = pd.Series([True, True, True, False, True, False],
index=['c1', 'c2', 'c3', 'c4', 'c5', 'c6'],
name='inclusion')
exp_metadata = pd.DataFrame(index=['c1', 'c2', 'c3', 'c4', 'c5', 'c6'])
exp_metadata.index.name = 'id'
exp_metadata = qiime2.Metadata(exp_metadata)
pdt.assert_series_equal(sel.inclusion, exp_inclusion)
self.assertEqual(sel.metadata, exp_metadata)
self.assertEqual(sel.label, 'sample_neighbors')
def test_sample_neighbors_metadata_superset(self):
context_md = self.get_data_path('context-metadata-2-extra-ids.tsv')
context_md = qiime2.Metadata.load(context_md)
sel = sample_neighbors(self.focal_seqs2,
self.context_seqs2,
percent_id=1.0,
samples_per_cluster=2,
locale=context_md.get_column('x'))
exp_inclusion = pd.Series([True], index=['c1'], name='inclusion')
exp_metadata = context_md.filter_ids(['c1'])
pdt.assert_series_equal(sel.inclusion, exp_inclusion)
self.assertEqual(sel.metadata, exp_metadata)
self.assertEqual(sel.label, 'sample_neighbors')
def test_sample_neighbors_locale(self):
count_obs_c2 = 0
count_obs_c3 = 0
count_obs_c4 = 0
count_obs_c5 = 0
exp_metadata = self.context_md1.to_dataframe()
exp_metadata.index.name = 'id'
exp_metadata = qiime2.Metadata(exp_metadata)
for _ in range(self._N_TEST_ITERATIONS):
sel = sample_neighbors(self.focal_seqs1,
self.context_seqs1,
percent_id=0.98,
samples_per_cluster=2,
locale=self.context_md1.get_column('x'))
obs_sampled_context_seqs = sel.inclusion[sel.inclusion].keys()
self.assertTrue('c1' in set(obs_sampled_context_seqs))
self.assertEqual(sel.inclusion.sum(), 3)
self.assertEqual(len(sel.inclusion), 6)
self.assertEqual(sel.metadata, exp_metadata)
self.assertEqual(sel.label, 'sample_neighbors')
if 'c2' in obs_sampled_context_seqs:
count_obs_c2 += 1
if 'c3' in obs_sampled_context_seqs:
count_obs_c3 += 1
if 'c4' in obs_sampled_context_seqs:
count_obs_c4 += 1
if 'c5' in obs_sampled_context_seqs:
count_obs_c5 += 1
# since c2, c3, and c5 all have locale "def" and c4 has locale "hijk",
# so we expect to see c4 more frequently than any of the other three
self.assertTrue(count_obs_c4 > count_obs_c2)
self.assertTrue(count_obs_c4 > count_obs_c3)
self.assertTrue(count_obs_c4 > count_obs_c5)
def test_sample_neighbors_invalid_max_accepts(self):
with self.assertRaisesRegex(ValueError, 'obtained per cluster'):
sample_neighbors(self.focal_seqs1,
self.context_seqs1,
percent_id=0.98,
samples_per_cluster=11)
