def test_mh_asymmetric_merge(track_abundance):
# test merging two asymmetric (different size) MHs
a = MinHash(20, 10, track_abundance=track_abundance)
for i in range(0, 40, 2):
a.add_hash(i)
# different size: 10
b = MinHash(10, 10, track_abundance=track_abundance)
for i in range(0, 80, 4):
b.add_hash(i)
c = a.merge(b)
d = b.merge(a)
assert len(a) == 20
assert len(b) == 10
assert len(c) == len(a)
assert len(d) == len(b)
# can't compare different sizes without downsampling
with pytest.raises(TypeError):
d.compare(a)
a = a.downsample_n(d.num)
print(a.get_mins())
print(d.get_mins())
assert d.compare(a) == 1.0
c = c.downsample_n(b.num)
assert c.compare(b) == 1.0
def test_mh_inplace_concat_asymmetric(track_abundance):
# test merging two asymmetric (different size) MHs
a = MinHash(20, 10, track_abundance=track_abundance)
for i in range(0, 40, 2):
a.add_hash(i)
# different size: 10
b = MinHash(10, 10, track_abundance=track_abundance)
for i in range(0, 80, 4):
b.add_hash(i)
c = a.__copy__()
c += b
d = b.__copy__()
d += a
assert len(a) == 20
assert len(b) == 10
assert len(c) == len(a)
assert len(d) == len(b)
try:
d.compare(a)
except TypeError as exc:
assert 'must have same num' in str(exc)
a = a.downsample_n(d.num)
assert d.compare(a) == 1.0 # see: d += a, above.
c = c.downsample_n(b.num)
assert c.compare(b) == 0.5
def test_mh_inplace_concat_asymmetric(track_abundance):
# test merging two asymmetric (different size) MHs
a = MinHash(20, 10, track_abundance=track_abundance)
for i in range(0, 40, 2):
a.add_hash(i)
# different size: 10
b = MinHash(10, 10, track_abundance=track_abundance)
for i in range(0, 80, 4):
b.add_hash(i)
c = a.__copy__()
c += b
d = b.__copy__()
d += a
assert len(a) == 20
assert len(b) == 10
assert len(c) == len(a)
assert len(d) == len(b)
try:
d.compare(a)
except TypeError as exc:
assert 'must have same num' in str(exc)
a = a.downsample_n(d.num)
assert d.compare(a) == 1.0 # see: d += a, above.
c = c.downsample_n(b.num)
assert c.compare(b) == 0.5
def test_mh_asymmetric_merge(track_abundance):
# test merging two asymmetric (different size) MHs
a = MinHash(20, 10, track_abundance=track_abundance)
for i in range(0, 40, 2):
a.add_hash(i)
# different size: 10
b = MinHash(10, 10, track_abundance=track_abundance)
for i in range(0, 80, 4):
b.add_hash(i)
c = a.merge(b)
d = b.merge(a)
assert len(a) == 20
assert len(b) == 10
assert len(c) == len(a)
assert len(d) == len(b)
# can't compare different sizes without downsampling
with pytest.raises(TypeError):
d.compare(a)
a = a.downsample_n(d.num)
print(a.get_mins())
print(d.get_mins())
assert d.compare(a) == 1.0
c = c.downsample_n(b.num)
assert c.compare(b) == 1.0
def test_mh_asymmetric(track_abundance):
a = MinHash(20, 10, track_abundance=track_abundance)
for i in range(0, 40, 2):
a.add_hash(i)
# different size: 10
b = MinHash(10, 10, track_abundance=track_abundance)
for i in range(0, 80, 4):
b.add_hash(i)
assert a.count_common(b) == 10
assert b.count_common(a) == 10
with pytest.raises(TypeError):
a.compare(b)
a = a.downsample_n(10)
assert a.compare(b) == 0.5
assert b.compare(a) == 0.5
def test_mh_asymmetric(track_abundance):
a = MinHash(20, 10, track_abundance=track_abundance)
for i in range(0, 40, 2):
a.add_hash(i)
# different size: 10
b = MinHash(10, 10, track_abundance=track_abundance)
for i in range(0, 80, 4):
b.add_hash(i)
assert a.count_common(b) == 10
assert b.count_common(a) == 10
with pytest.raises(TypeError):
a.compare(b)
a = a.downsample_n(10)
assert a.compare(b) == 0.5
assert b.compare(a) == 0.5
def test_mh_jaccard_asymmetric_num(track_abundance):
a = MinHash(20, 10, track_abundance=track_abundance)
for i in range(0, 40, 2):
a.add_hash(i)
# different size: 10
b = MinHash(10, 10, track_abundance=track_abundance)
for i in range(0, 80, 4):
b.add_hash(i)
assert a.count_common(b) == 10
assert b.count_common(a) == 10
# with 'jaccard', this will raise an error b/c different num
with pytest.raises(TypeError):
a.jaccard(b)
a = a.downsample_n(10)
# CTB note: this used to be 'compare', is now 'jaccard'
assert a.jaccard(b) == 0.5
assert b.jaccard(a) == 0.5
def test_mh_asymmetric_merge(track_abundance):
# test merging two asymmetric (different size) MHs
a = MinHash(20, 10, track_abundance=track_abundance)
for i in range(0, 40, 2):
a.add_hash(i)
# different size: 10
b = MinHash(10, 10, track_abundance=track_abundance)
for i in range(0, 80, 4):
b.add_hash(i)
c = a.merge(b)
d = b.merge(a)
assert len(a) == 20
assert len(b) == 10
assert len(c) == len(a)
assert len(d) == len(b)
# can't use jaccard on different nums without downsampling
with pytest.raises(TypeError):
d.jaccard(a)
a = a.downsample_n(d.num)
print(a.get_mins())
print(d.get_mins())
if track_abundance:
assert round(d.similarity(a), 3) == 0.91
else:
assert round(d.similarity(a), 3) == 1.0
c = c.downsample_n(b.num)
if track_abundance:
assert round(c.similarity(b), 3) == 0.91
else:
assert c.similarity(b) == 1.0
def test_mh_downsample_n_error(track_abundance):
a = MinHash(20, 10, track_abundance=track_abundance)
with pytest.raises(ValueError):
a.downsample_n(30)
