class TestHdHash(TestCase):
nr_input_bits = 32 * 5
nr_output_bits = 8
h = hdhash.HdHash(nr_input_bits, nr_output_bits)
def test_hashing_with_zeros(self):
#givne
data = [0] * 5
#when
value = self.h.hashing(data)
#then
self.assertEqual(0, value)
def test_hashing_with_ones(self):
#givne
data = [0xffffffff] * 5
#when
value = self.h.hashing(data)
#then
self.assertEqual(0xff, value)
def test_hashing_with_big_input(self):
#given
data = [0x0] * 10
#when
value = self.h.hashing(data)
#then
self.assertEqual(0, value)
class TestLshLookupTable(TestCase):
hash = hdhash.HdHash(WINDOW_SIZE / BAND * NR_BIT_PER_WORD, 8)
l = lsh.Lsh(WINDOW_SIZE, BAND, ROW, hash)
def test_one_signature(self):
#given
signature = [0x0]
#when
lookup = self.l.lookup_table(signature)
#then
self.assertEqual(1, len(lookup))
self.assertEqual(['s0b0', [-1]], lookup[0])
return
def test_n_signature(self):
#given
signature = [0xff, 0]
#when
lookup = self.l.lookup_table(signature)
#then
self.assertEqual(2, len(lookup))
return
def test_n_band_signature(self):
#given
nr_unique_signature = 9
signature = list(range(nr_unique_signature)) * BAND
#when
lookup = self.l.lookup_table(signature)
#then
self.assertEqual(nr_unique_signature * BAND, len(lookup))
def test_lookup_table_with_None(self):
#given
signature = None
#when
lookup = self.l.lookup_table(signature)
#then
self.assertEqual([], lookup)
return
def test_lookup_tabke_with_empty(self):
#given
signature = []
#when
lookup = self.l.lookup_table(signature)
#then
self.assertEqual([], lookup)
class TestLshHashing(TestCase):
hash = hdhash.HdHash(WINDOW_SIZE / BAND * NR_BIT_PER_WORD, 8)
l = lsh.Lsh(WINDOW_SIZE, BAND, ROW, hash)
def test_hashing(self):
#given
data = [0] * WINDOW_SIZE * 2
#when
value = self.l.hashing(data)
#then
self.assertEqual(BAND * 2, len(value))
self.assertEqual('0' * BAND * 2, "".join(str(x) for x in value))
class TestParallelSearch(TestCase):
window_size = 155
band = 31
row = 8
band_size = int(window_size / band)
sc = SparkContext(appName='SearchMusic')
ss = SparkSession.builder.appName('SearchMusic').master(
'local[*]').getOrCreate()
query_df, music_df = music_query_parser.do(sc, ss, query_file, music_file)
h = hdhash.HdHash(band_size * NR_BIT_PER_WORD, row)
search = parallel_search.ParallelSearch(ss, query_df, music_df,
window_size, band, row, h,
hdhash.hamming_distance)
def test_by_qid_mid(self):
self.search.by_qid_and_mid(21, 42, 0.35)
class TestParallelLsh(TestCase):
window_size = 155
band = 31
row = 8
band_size = int(window_size/band)
sc = SparkContext(appName='SearchMusic')
ss = SparkSession.builder.appName('SearchMusic').master('local[*]').getOrCreate()
query_df, music_df = music_query_parser.do(sc, ss, query_file, music_file)
h = hdhash.HdHash(band_size * NR_BIT_PER_WORD, row)
search = parallel_search.ParallelSearch(ss, query_df, music_df, window_size, band, row, h, hdhash.hamming_distance)
def test_lookup_table(self):
mdf = self.music_df
mdf = self.p.lookup_table(mdf, 'mcode', 'mid')
mdf.persist()
mdf.show()
sc, ss = spark_init()
window_size = 155
band = 31
row = 8
band_size = int(window_size/band)
t1 = time.time()
query_df, music_df = music_query_parser.do(sc, ss, qfile, mfile)
#debug
print ('befor search')
lsh_hash = hdhash.HdHash(band_size * NR_BIT_PER_WORD, row)
search = parallel_search.ParallelSearch(ss, query_df, music_df,
window_size, band, row,
lsh_hash, hdhash.hamming_distance)
t2 = time.time()
# qid = 7
# searched = search.by_qid(qid, THRESHOLD)
# if not searched:
# print ('Search failed: Query[%d]' % qid)
# else:
# for s in searched:
# print ('Search success: Query[%d] hit on %s, songid(%d), code_idx(%d)~code_idx(%d)' %(s['qid'], s['file'], s['mid'], s['col']*5, s['col']*5 + window_size))
nr_query = 30