def test_simhash128_3(self): sh = SimHashSignature(128) str1 = "♡♥❤❥" str2 = u"♡♥❤❥" sig5 = sh.get_signature(str1) sig6 = sh.get_signature(str2) self.assertNotEqual(sig5, sig6)
def test_simhash_similarity(self): """Signatures should be consistent""" sh = SimHashSignature(64) sig1 = sh.get_signature("abracadabra") sig2 = sh.get_signature("") dist = hamming(sig1, sig2) self.assertEqual(dist, 25) sig1 = sh.get_signature("abracadabra") sig2 = sh.get_signature("abracadabra") dist = hamming(sig1, sig2) self.assertEqual(dist, 0) sig1 = sh.get_signature("abracadabra") sig2 = sh.get_signature("abracdabra") dist = hamming(sig1, sig2) self.assertEqual(dist, 6)
def test_simhash_similarity(self): """Signatures should be consistent""" sh = SimHashSignature(64) sig1 = sh.get_signature("abracadabra") sig2 = sh.get_signature("abracadabra") dist = hamming(sig1, sig2) self.assertEqual(dist, 0) sig1 = sh.get_signature("abracadabra") sig2 = sh.get_signature("arbcd") dist = hamming(sig1, sig2) self.assertEqual(dist, 13) sig1 = sh.get_signature("abracadabra") sig2 = sh.get_signature("") dist = hamming(sig1, sig2) self.assertEqual(dist, 32)
def test_simhash64(self): """Simhash signature of an empty string should be zero and unicode and regular strings should give the same simhash signatures """ sh = SimHashSignature(64) sig1 = sh.get_signature("") sig2 = sh.get_signature(u"") self.assertEqual(sig1, sig2) sig3 = sh.get_signature("abracadabra") sig4 = sh.get_signature(u"abracadabra") self.assertEqual(sig3, sig4) str1 = "♡♥❤❥" str2 = u"♡♥❤❥" sig5 = sh.get_signature(str1) sig6 = sh.get_signature(str2) self.assertNotEqual(sig5, sig6)
def test_simhash128(self): """Simhash signature of an empty string should be zero and unicode and regular strings should give the same simhash signatures """ sh = SimHashSignature(128) sig1 = sh.get_signature("") sig2 = sh.get_signature(u"") self.assertEqual(sig1, 0) self.assertEqual(sig1, sig2) sig3 = sh.get_signature("abracadabra") sig4 = sh.get_signature(u"abracadabra") self.assertEqual(sig3, sig4) str1 = "♡♥❤❥" str2 = u"♡♥❤❥" sig5 = sh.get_signature(str1) sig6 = sh.get_signature(str2) self.assertNotEqual(sig5, sig6)
def test_simhash_feature_weights(self): """Features should be weighted and should contribute to similarity/distance """ sh = SimHashSignature(64) sig1 = sh.get_signature("abracadabra") sig2 = sh.get_signature("abracdabra") dist = hamming(sig1, sig2) self.assertEqual(dist, 6) sig1 = sh.get_signature("abracadabra", ("cats", 0)) sig2 = sh.get_signature("abracdabra", ("dogs", 0)) dist = hamming(sig1, sig2) self.assertEqual(dist, 6) sig1 = sh.get_signature("abracadabra", ("cats", 0)) sig2 = sh.get_signature("abracadabra", ("dogs", 0)) dist = hamming(sig1, sig2) self.assertEqual(dist, 0) sig1 = sh.get_signature("abracadabra", ("ca", 4)) sig2 = sh.get_signature("abracadabra", ("do", 4)) dist = hamming(sig1, sig2) self.assertEqual(dist, 0) sig1 = sh.get_signature("abracadabra", ("ca", 5)) sig2 = sh.get_signature("abracadabra", ("do", 5)) dist = hamming(sig1, sig2) self.assertEqual(dist, 7) sig1 = sh.get_signature("abracadabra", ("cats", 200)) sig2 = sh.get_signature("abracadabra", ("dogs", 200)) dist = hamming(sig1, sig2) self.assertEqual(dist, 17) sig1 = sh.get_signature("abracadabra", ("cats", 10)) sig2 = sh.get_signature("abracadabra", ("cats", 10)) dist = hamming(sig1, sig2) self.assertEqual(dist, 0)
def test_simhash_feature_weights(self): """Features should be weighted and should contribute to similarity/distance """ sh = SimHashSignature(64) sig1 = sh.get_signature("abracadabra") sig2 = sh.get_signature("abracdabra") dist = hamming(sig1, sig2) self.assertEqual(dist, 3) sig1 = sh.get_signature("abracadabra", ("cats", 0)) sig2 = sh.get_signature("abracdabra", ("dogs", 0)) dist = hamming(sig1, sig2) self.assertEqual(dist, 3) sig1 = sh.get_signature("abracadabra", ("cats", 0)) sig2 = sh.get_signature("abracadabra", ("dogs", 0)) dist = hamming(sig1, sig2) self.assertEqual(dist, 0) sig1 = sh.get_signature("abracadabra", ("ca", 4)) sig2 = sh.get_signature("abracadabra", ("do", 4)) dist = hamming(sig1, sig2) self.assertEqual(dist, 9) sig1 = sh.get_signature("abracadabra", ("ca", 5)) sig2 = sh.get_signature("abracadabra", ("do", 5)) dist = hamming(sig1, sig2) self.assertEqual(dist, 9) sig1 = sh.get_signature("abracadabra", ("cats", 200)) sig2 = sh.get_signature("abracadabra", ("dogs", 200)) dist = hamming(sig1, sig2) self.assertEqual(dist, 21) sig1 = sh.get_signature("abracadabra", ("cats", 10)) sig2 = sh.get_signature("abracadabra", ("cats", 10)) dist = hamming(sig1, sig2) self.assertEqual(dist, 0)
def test_simhash_similarity_3(self): sh = SimHashSignature(64) sig1 = sh.get_signature("abracadabra") sig2 = sh.get_signature("") dist = hamming(sig1, sig2) self.assertEqual(37, dist)
def test_simhash128_2(self): sh = SimHashSignature(128) sig3 = sh.get_signature("abracadabra") sig4 = sh.get_signature(u"abracadabra") self.assertEqual(sig3, sig4)
def test_simhash128_1(self): sh = SimHashSignature(128) sig1 = sh.get_signature("") sig2 = sh.get_signature(u"") self.assertEqual(0, sig1) self.assertEqual(sig1, sig2)
def test_simhash64_1(self): sh = SimHashSignature(64) sig1 = sh.get_signature("") sig2 = sh.get_signature(u"") self.assertEqual(sig1, sig2)
def test_simhash_feature_weights_7(self): sh = SimHashSignature(64) sig1 = sh.get_signature("abracadabra", ("cats", 10)) sig2 = sh.get_signature("abracadabra", ("cats", 10)) dist = hamming(sig1, sig2) self.assertEqual(0, dist)
def test_simhash_feature_weights_1(self): sh = SimHashSignature(64) sig1 = sh.get_signature("abracadabra") sig2 = sh.get_signature("abracdabra") dist = hamming(sig1, sig2) self.assertEqual(3, dist)