class DissimilarityMatrixTests(DissimilarityMatrixTestData):
def setUp(self):
super(DissimilarityMatrixTests, self).setUp()
self.dm_1x1 = DissimilarityMatrix(self.dm_1x1_data, ['a'])
self.dm_2x2 = DissimilarityMatrix(self.dm_2x2_data, ['a', 'b'])
self.dm_2x2_asym = DissimilarityMatrix(self.dm_2x2_asym_data,
['a', 'b'])
self.dm_3x3 = DissimilarityMatrix(self.dm_3x3_data, ['a', 'b', 'c'])
self.dms = [self.dm_1x1, self.dm_2x2, self.dm_2x2_asym, self.dm_3x3]
self.dm_shapes = [(1, 1), (2, 2), (2, 2), (3, 3)]
self.dm_sizes = [1, 4, 4, 9]
self.dm_transposes = [
self.dm_1x1, self.dm_2x2,
DissimilarityMatrix([[0, -2], [1, 0]], ['a', 'b']), self.dm_3x3]
self.dm_redundant_forms = [np.array(self.dm_1x1_data),
np.array(self.dm_2x2_data),
np.array(self.dm_2x2_asym_data),
np.array(self.dm_3x3_data)]
def test_io(self):
# Very basic check that read/write public API is present and appears to
# be functioning. Roundtrip from memory -> disk -> memory and ensure
# results match.
fh = StringIO()
self.dm_3x3.write(fh)
fh.seek(0)
deserialized = DissimilarityMatrix.read(fh)
self.assertEqual(deserialized, self.dm_3x3)
self.assertTrue(type(deserialized) == DissimilarityMatrix)
def test_deprecated_io(self):
fh = StringIO()
npt.assert_warns(UserWarning, self.dm_3x3.to_file, fh)
fh.seek(0)
deserialized = npt.assert_warns(UserWarning,
DissimilarityMatrix.from_file, fh)
self.assertEqual(deserialized, self.dm_3x3)
self.assertTrue(type(deserialized) == DissimilarityMatrix)
def test_init_from_dm(self):
ids = ['foo', 'bar', 'baz']
# DissimilarityMatrix -> DissimilarityMatrix
exp = DissimilarityMatrix(self.dm_3x3_data, ids)
obs = DissimilarityMatrix(self.dm_3x3, ids)
self.assertEqual(obs, exp)
# Test that copy of data is not made.
self.assertTrue(obs.data is self.dm_3x3.data)
obs.data[0, 1] = 424242
self.assertTrue(np.array_equal(obs.data, self.dm_3x3.data))
# DistanceMatrix -> DissimilarityMatrix
exp = DissimilarityMatrix(self.dm_3x3_data, ids)
obs = DissimilarityMatrix(
DistanceMatrix(self.dm_3x3_data, ('a', 'b', 'c')), ids)
self.assertEqual(obs, exp)
# DissimilarityMatrix -> DistanceMatrix
with self.assertRaises(DistanceMatrixError):
DistanceMatrix(self.dm_2x2_asym, ['foo', 'bar'])
def test_init_no_ids(self):
exp = DissimilarityMatrix(self.dm_3x3_data, ('0', '1', '2'))
obs = DissimilarityMatrix(self.dm_3x3_data)
self.assertEqual(obs, exp)
self.assertEqual(obs['1', '2'], 12.0)
def test_init_invalid_input(self):
# Empty data.
with self.assertRaises(DissimilarityMatrixError):
DissimilarityMatrix([], [])
# Another type of empty data.
with self.assertRaises(DissimilarityMatrixError):
DissimilarityMatrix(np.empty((0, 0)), [])
# Invalid number of dimensions.
with self.assertRaises(DissimilarityMatrixError):
DissimilarityMatrix([1, 2, 3], ['a'])
# Dimensions don't match.
with self.assertRaises(DissimilarityMatrixError):
DissimilarityMatrix([[1, 2, 3]], ['a'])
data = [[0, 1], [1, 0]]
# Duplicate IDs.
with self.assertRaises(DissimilarityMatrixError):
DissimilarityMatrix(data, ['a', 'a'])
# Number of IDs don't match dimensions.
with self.assertRaises(DissimilarityMatrixError):
DissimilarityMatrix(data, ['a', 'b', 'c'])
# Non-hollow.
data = [[0.0, 1.0], [1.0, 0.01]]
with self.assertRaises(DissimilarityMatrixError):
DissimilarityMatrix(data, ['a', 'b'])
def test_data(self):
for dm, exp in zip(self.dms, self.dm_redundant_forms):
obs = dm.data
self.assertTrue(np.array_equal(obs, exp))
with self.assertRaises(AttributeError):
self.dm_3x3.data = 'foo'
def test_ids(self):
obs = self.dm_3x3.ids
self.assertEqual(obs, ('a', 'b', 'c'))
# Test that we overwrite the existing IDs and that the ID index is
# correctly rebuilt.
new_ids = ['foo', 'bar', 'baz']
self.dm_3x3.ids = new_ids
obs = self.dm_3x3.ids
self.assertEqual(obs, tuple(new_ids))
self.assertTrue(np.array_equal(self.dm_3x3['bar'],
np.array([0.01, 0.0, 12.0])))
with self.assertRaises(MissingIDError):
self.dm_3x3['b']
def test_ids_invalid_input(self):
with self.assertRaises(DissimilarityMatrixError):
self.dm_3x3.ids = ['foo', 'bar']
# Make sure that we can still use the dissimilarity matrix after trying
# to be evil.
obs = self.dm_3x3.ids
self.assertEqual(obs, ('a', 'b', 'c'))
def test_dtype(self):
for dm in self.dms:
self.assertEqual(dm.dtype, np.float64)
def test_shape(self):
for dm, shape in zip(self.dms, self.dm_shapes):
self.assertEqual(dm.shape, shape)
def test_size(self):
for dm, size in zip(self.dms, self.dm_sizes):
self.assertEqual(dm.size, size)
def test_transpose(self):
for dm, transpose in zip(self.dms, self.dm_transposes):
self.assertEqual(dm.T, transpose)
self.assertEqual(dm.transpose(), transpose)
# We should get a reference to a different object back, even if the
# transpose is the same as the original.
self.assertTrue(dm.transpose() is not dm)
def test_index(self):
self.assertEqual(self.dm_3x3.index('a'), 0)
self.assertEqual(self.dm_3x3.index('b'), 1)
self.assertEqual(self.dm_3x3.index('c'), 2)
with self.assertRaises(MissingIDError):
self.dm_3x3.index('d')
with self.assertRaises(MissingIDError):
self.dm_3x3.index(1)
def test_redundant_form(self):
for dm, redundant in zip(self.dms, self.dm_redundant_forms):
obs = dm.redundant_form()
self.assertTrue(np.array_equal(obs, redundant))
def test_copy(self):
copy = self.dm_2x2.copy()
self.assertEqual(copy, self.dm_2x2)
self.assertFalse(copy.data is self.dm_2x2.data)
# deepcopy doesn't actually create a copy of the IDs because it is a
# tuple of strings, which is fully immutable.
self.assertTrue(copy.ids is self.dm_2x2.ids)
new_ids = ['hello', 'world']
copy.ids = new_ids
self.assertNotEqual(copy.ids, self.dm_2x2.ids)
copy = self.dm_2x2.copy()
copy.data[0, 1] = 0.0001
self.assertFalse(np.array_equal(copy.data, self.dm_2x2.data))
def test_filter_no_filtering(self):
# Don't actually filter anything -- ensure we get back a different
# object.
obs = self.dm_3x3.filter(['a', 'b', 'c'])
self.assertEqual(obs, self.dm_3x3)
self.assertFalse(obs is self.dm_3x3)
def test_filter_reorder(self):
# Don't filter anything, but reorder the distance matrix.
order = ['c', 'a', 'b']
exp = DissimilarityMatrix(
[[0, 4.2, 12], [4.2, 0, 0.01], [12, 0.01, 0]], order)
obs = self.dm_3x3.filter(order)
self.assertEqual(obs, exp)
def test_filter_single_id(self):
ids = ['b']
exp = DissimilarityMatrix([[0]], ids)
obs = self.dm_2x2_asym.filter(ids)
self.assertEqual(obs, exp)
def test_filter_asymmetric(self):
# 2x2
ids = ['b', 'a']
exp = DissimilarityMatrix([[0, -2], [1, 0]], ids)
obs = self.dm_2x2_asym.filter(ids)
self.assertEqual(obs, exp)
# 3x3
dm = DissimilarityMatrix([[0, 10, 53], [42, 0, 22.5], [53, 1, 0]],
('bro', 'brah', 'breh'))
ids = ['breh', 'brah']
exp = DissimilarityMatrix([[0, 1], [22.5, 0]], ids)
obs = dm.filter(ids)
self.assertEqual(obs, exp)
def test_filter_subset(self):
ids = ('c', 'a')
exp = DissimilarityMatrix([[0, 4.2], [4.2, 0]], ids)
obs = self.dm_3x3.filter(ids)
self.assertEqual(obs, exp)
ids = ('b', 'a')
exp = DissimilarityMatrix([[0, 0.01], [0.01, 0]], ids)
obs = self.dm_3x3.filter(ids)
self.assertEqual(obs, exp)
# 4x4
dm = DissimilarityMatrix([[0, 1, 55, 7], [1, 0, 16, 1],
[55, 16, 0, 23], [7, 1, 23, 0]])
ids = np.asarray(['3', '0', '1'])
exp = DissimilarityMatrix([[0, 7, 1], [7, 0, 1], [1, 1, 0]], ids)
obs = dm.filter(ids)
self.assertEqual(obs, exp)
def test_filter_duplicate_ids(self):
with self.assertRaises(DissimilarityMatrixError):
self.dm_3x3.filter(['c', 'a', 'c'])
def test_filter_missing_ids(self):
with self.assertRaises(MissingIDError):
self.dm_3x3.filter(['c', 'bro'])
def test_filter_missing_ids_strict_false(self):
# no exception should be raised
ids = ('c', 'a')
exp = DissimilarityMatrix([[0, 4.2], [4.2, 0]], ids)
obs = self.dm_3x3.filter(['c', 'a', 'not found'], strict=False)
self.assertEqual(obs, exp)
def test_filter_empty_ids(self):
with self.assertRaises(DissimilarityMatrixError):
self.dm_3x3.filter([])
def test_str(self):
for dm in self.dms:
obs = str(dm)
# Do some very light testing here to make sure we're getting a
# non-empty string back. We don't want to test the exact
# formatting.
self.assertTrue(obs)
def test_eq(self):
for dm in self.dms:
copy = dm.copy()
self.assertTrue(dm == dm)
self.assertTrue(copy == copy)
self.assertTrue(dm == copy)
self.assertTrue(copy == dm)
self.assertFalse(self.dm_1x1 == self.dm_3x3)
def test_ne(self):
# Wrong class.
self.assertTrue(self.dm_3x3 != 'foo')
# Wrong shape.
self.assertTrue(self.dm_3x3 != self.dm_1x1)
# Wrong IDs.
other = self.dm_3x3.copy()
other.ids = ['foo', 'bar', 'baz']
self.assertTrue(self.dm_3x3 != other)
# Wrong data.
other = self.dm_3x3.copy()
other.data[1, 0] = 42.42
self.assertTrue(self.dm_3x3 != other)
self.assertFalse(self.dm_2x2 != self.dm_2x2)
def test_contains(self):
self.assertTrue('a' in self.dm_3x3)
self.assertTrue('b' in self.dm_3x3)
self.assertTrue('c' in self.dm_3x3)
self.assertFalse('d' in self.dm_3x3)
def test_getslice(self):
# Slice of first dimension only. Test that __getslice__ defers to
# __getitem__.
obs = self.dm_2x2[1:]
self.assertTrue(np.array_equal(obs, np.array([[0.123, 0.0]])))
self.assertEqual(type(obs), np.ndarray)
def test_getitem_by_id(self):
obs = self.dm_1x1['a']
self.assertTrue(np.array_equal(obs, np.array([0.0])))
obs = self.dm_2x2_asym['b']
self.assertTrue(np.array_equal(obs, np.array([-2.0, 0.0])))
obs = self.dm_3x3['c']
self.assertTrue(np.array_equal(obs, np.array([4.2, 12.0, 0.0])))
with self.assertRaises(MissingIDError):
self.dm_2x2['c']
def test_getitem_by_id_pair(self):
# Same object.
self.assertEqual(self.dm_1x1['a', 'a'], 0.0)
# Different objects (symmetric).
self.assertEqual(self.dm_3x3['b', 'c'], 12.0)
self.assertEqual(self.dm_3x3['c', 'b'], 12.0)
# Different objects (asymmetric).
self.assertEqual(self.dm_2x2_asym['a', 'b'], 1.0)
self.assertEqual(self.dm_2x2_asym['b', 'a'], -2.0)
with self.assertRaises(MissingIDError):
self.dm_2x2['a', 'c']
def test_getitem_ndarray_indexing(self):
# Single element access.
obs = self.dm_3x3[0, 1]
self.assertEqual(obs, 0.01)
# Single element access (via two __getitem__ calls).
obs = self.dm_3x3[0][1]
self.assertEqual(obs, 0.01)
# Row access.
obs = self.dm_3x3[1]
self.assertTrue(np.array_equal(obs, np.array([0.01, 0.0, 12.0])))
self.assertEqual(type(obs), np.ndarray)
# Grab all data.
obs = self.dm_3x3[:, :]
self.assertTrue(np.array_equal(obs, self.dm_3x3.data))
self.assertEqual(type(obs), np.ndarray)
with self.assertRaises(IndexError):
self.dm_3x3[:, 3]
def test_validate_invalid_dtype(self):
with self.assertRaises(DissimilarityMatrixError):
self.dm_3x3._validate(np.array([[0, 42], [42, 0]]), ['a', 'b'])
def test_pprint_ids(self):
# No truncation.
exp = 'a, b, c'
obs = self.dm_3x3._pprint_ids()
self.assertEqual(obs, exp)
# Truncation.
exp = 'a, b, ...'
obs = self.dm_3x3._pprint_ids(max_chars=5)
self.assertEqual(obs, exp)
dm = np.zeros([len(samples),len(samples)])
pm = np.zeros([len(samples),len(samples)])
# fill matrices with values
for s1, s2, d, p in zip(mash_vec[0],mash_vec[1],mash_vec[2],mash_vec[3]):
i1 = samples.index(s1)
i2 = samples.index(s2)
print('s1: %s, s2: %s, i1: %s, i2: %s, d: %s, p: %s' % (s1, s2, i1, i2, d, p))
dm[i1,i2] = d
dm[i2,i1] = d
pm[i1,i2] = p
pm[i2,i1] = p
ids = [os.path.basename(x) for x in samples]
sk_dm = DissimilarityMatrix(dm, ids=ids)
sk_pm = DissimilarityMatrix(pm, ids=ids)
sk_dm.write(output['dist_matrix'])
sk_pm.write(output['p_matrix'])
#### Mash rules
rule mash:
input:
expand(mash_dir + '{sample}/mash/{sample}.msh',
sample=samples),
expand(mash_dir + '{sample}/mash/{sample}.refseq.txt',
sample=samples),
mash_dir + 'combined_analysis/mash.dist.dm',
mash_dir + 'combined_analysis/mash.dist.p'
