def test_valid(self):
inputs = [((190, 1, 0), (190, 1, 0), 0),
((190, 1, 0), (190, 1, 1), 1 / 3),
((190, 0, 1), (190, 1, 0), 2 / 3),
((190, 0, 1), (191, 1, 0), 1),
((190, 0, 1), (190, 0, 1, 0), 1 / 4), ((190, ), (), 1),
((), (), 1)]
for left, right, result in inputs:
self.assertTrue(
float_cmp(compare_single_path(left, right), result)
and float_cmp(compare_single_path(right, left), result))
def test_invalid(self):
'Test rather unusual corner cases'
calc = GenreTreeDistance(GenreTreeProvider())
self.assertTrue(float_cmp(calc.compute([], []), 1.0))
self.assertTrue(float_cmp(calc.compute([], [(1, 0)]), 1.0))
self.assertTrue(float_cmp(calc.compute([], ['berta']), 1.0))
# Funny one (strings are iterable)
self.assertTrue(float_cmp(calc.compute(['berta'], ['berta']), 0.0))
# Passing a non-iterable:
with self.assertRaises(TypeError):
calc.compute([1], [2])
def test_invalid(self):
'Test rather unusual corner cases'
calc = GenreTreeDistance(GenreTreeProvider())
self.assertTrue(float_cmp(calc.compute([], []), 1.0))
self.assertTrue(float_cmp(calc.compute([], [(1, 0)]), 1.0))
self.assertTrue(float_cmp(calc.compute([], ['berta']), 1.0))
# Funny one (strings are iterable)
self.assertTrue(float_cmp(calc.compute(['berta'], ['berta']), 0.0))
# Passing a non-iterable:
with self.assertRaises(TypeError):
calc.compute([1], [2])
def test_valid(self):
inputs = [
((190, 1, 0), (190, 1, 0), 0),
((190, 1, 0), (190, 1, 1), 1 / 3),
((190, 0, 1), (190, 1, 0), 2 / 3),
((190, 0, 1), (191, 1, 0), 1),
((190, 0, 1), (190, 0, 1, 0), 1 / 4),
((190, ), (), 1),
((), (), 1)
]
for left, right, result in inputs:
self.assertTrue(
float_cmp(compare_single_path(left, right), result)
and
float_cmp(compare_single_path(right, left), result)
)
def test_weight(self):
dist = Distance(self._session, {'genre': 1.0})
self.assertAlmostEqual(dist.distance, 1.0)
dist = Distance(self._session, {'random': 1.0})
self.assertAlmostEqual(dist.distance, 1.0)
dist = Distance(self._session, {'genre': 1.0, 'random': 1.0})
self.assertAlmostEqual(dist.distance, 1.0)
dist = Distance(self._session, {'genre': 1.0, 'random': 0.0})
self.assertAlmostEqual(dist.distance, 5 / 6)
dist = Distance(self._session, {'genre': 0.0, 'random': 0.0})
self.assertAlmostEqual(dist.distance, 0.0)
# Compute it manually:
dist = Distance(self._session, {'genre': 0.5, 'random': 0.1})
self.assertTrue(float_cmp(dist.distance, (0.5 * 0.5 + 0.1 * 0.1) / 0.6))
def do_compute(self, lefts, rights):
"""Calculate distance between two genre paths by using complete linkage.
:param lefts: A list of Genre Paths.
:param rights: A list of Genre Paths to compare with.
:returns: A distance between 0.0 and 1.0 (max diversity.)
"""
min_dist = 1.0
for left, right in product(lefts, rights):
min_dist = min(min_dist, compare_single_path(left, right))
# Optimization: Often we get a low value early.
if float_cmp(min_dist, 0.0):
break
return min_dist
def test_weight(self):
dist = Distance(self._session, {'genre': 1.0})
self.assertAlmostEqual(dist.distance, 1.0)
dist = Distance(self._session, {'random': 1.0})
self.assertAlmostEqual(dist.distance, 1.0)
dist = Distance(self._session, {'genre': 1.0, 'random': 1.0})
self.assertAlmostEqual(dist.distance, 1.0)
dist = Distance(self._session, {'genre': 1.0, 'random': 0.0})
self.assertAlmostEqual(dist.distance, 5 / 6)
dist = Distance(self._session, {'genre': 0.0, 'random': 0.0})
self.assertAlmostEqual(dist.distance, 0.0)
# Compute it manually:
dist = Distance(self._session, {'genre': 0.5, 'random': 0.1})
self.assertTrue(
float_cmp(dist.distance, (0.5 * 0.5 + 0.1 * 0.1) / 0.6))
def do_compute(self, lefts, rights):
"""Calculate distance between two genre paths by using complete linkage.
:param lefts: A list of Genre Paths.
:param rights: A list of Genre Paths to compare with.
:returns: A distance between 0.0 and 1.0 (max diversity.)
"""
min_dist = 1.0
for left, right in product(lefts, rights):
min_dist = min(min_dist, compare_single_path(left, right))
# Optimization: Often we get a low value early.
if float_cmp(min_dist, 0.0):
break
return min_dist
def do_compute(self, lefts, rights):
if not lefts or not rights:
return 1.0
left_lang, lefts = lefts
right_lang, rights = rights
if right_lang != left_lang:
return 1.0
min_distance = 1.0
for kwa, kwb in product(lefts, rights):
# common = sum(1 for _ in takewhile(lambda t: t[0] == t[1], zip(kwa, kwb)))
distance = 1.0 - len(kwa & kwb) / max(4, max(len(kwa), len(kwb)))
min_distance = min(distance, min_distance)
if float_cmp(distance, 0.0):
break
return 0.67 * min_distance + 0.33 * (not right_lang == left_lang)
def do_compute(self, lefts, rights):
if not lefts or not rights:
return 1.0
left_lang, lefts = lefts
right_lang, rights = rights
if right_lang != left_lang:
return 1.0
min_distance = 1.0
for kwa, kwb in product(lefts, rights):
# common = sum(1 for _ in takewhile(lambda t: t[0] == t[1], zip(kwa, kwb)))
distance = 1.0 - len(kwa & kwb) / max(4, max(len(kwa), len(kwb)))
min_distance = min(distance, min_distance)
if float_cmp(distance, 0.0):
break
return 0.67 * min_distance + 0.33 * (not right_lang == left_lang)
def full_cross_compare(expected):
self.assertTrue(float_cmp(calc.compute(a, b), expected))
self.assertTrue(float_cmp(calc.compute(b, a), expected))
self.assertTrue(float_cmp(calc.compute(a, a), 0.0))
self.assertTrue(float_cmp(calc.compute(b, b), 0.0))
def __eq__(self, other):
return float_cmp(self.distance, other.distance)
def __eq__(self, other):
return float_cmp(self.distance, other.distance)
def full_cross_compare(expected):
self.assertTrue(float_cmp(calc.compute(a, b), expected))
self.assertTrue(float_cmp(calc.compute(b, a), expected))
self.assertTrue(float_cmp(calc.compute(a, a), 0.0))
self.assertTrue(float_cmp(calc.compute(b, b), 0.0))
