def _vid_for_value(self, column, key):
if column.is_trie:
if column.rtrie_indicator == mdb.MDB_UINT_16:
key = rtrie.vid_for_value(self.vid16_nodes, self.vid16_kids, key)
else:
key = rtrie.vid_for_value(self.vid_nodes, self.vid_kids, key)
elif column.is_lz4:
key = clz4.compress(key)
return key
def _vid_for_value(self, column, key):
if column.is_trie:
if column.rtrie_indicator == mdb.MDB_UINT_16:
key = rtrie.vid_for_value(self.vid16_nodes, self.vid16_kids, key)
else:
key = rtrie.vid_for_value(self.vid_nodes, self.vid_kids, key)
elif column.is_lz4:
key = clz4.compress(key)
return key
def test_marble_insert(self):
# test general infomation
self.assertEqual(self.n_inserted, len(_ALBUMS))
self.assertEqual(_NPARTITIONS, len(self.files))
part_id = {}
# test that each sub db is fine
for date, file in self.files.iteritems():
env, txn, dbs, meta = self.marble._open(file)
# check meta db
self.assertTrue(meta.contains(txn, "_vid_nodes"))
self.assertTrue(meta.contains(txn, "_vid_kids"))
self.assertTrue(meta.contains(txn, "_vid16_nodes"))
self.assertTrue(meta.contains(txn, "_vid16_kids"))
self.assertEqual(meta.get(txn, "name"), ujson.dumps("Collections"))
self.assertEqual(meta.get(txn, "partition"), ujson.dumps("date"))
self.assertEqual(meta.get(txn, "fields"), ujson.dumps(_FIELDS))
self.assertEqual(meta.get(txn, "_pdata"), ujson.dumps(date))
vid_nodes, _ = meta.get_raw(txn, '_vid_nodes')
vid_kids, _ = meta.get_raw(txn, '_vid_kids')
vid16_nodes, _ = meta.get_raw(txn, '_vid16_nodes', (None, 0))
vid16_kids, _ = meta.get_raw(txn, '_vid16_kids', (None, 0))
# check subdb, subinddb
part_id[date] = 1
for name, (db, ind_db, _, column, _) in dbs.iteritems():
if name == "_count":
continue
bitmaps = {}
part_id[date] = 1
for album in self.albums:
if date == album[_PARTITIONS]: # match the partition
value = album[name]
i = part_id[album[_PARTITIONS]]
part_id[album[_PARTITIONS]] += 1
if column.is_trie:
if column.rtrie_indicator == mdb.MDB_UINT_16:
val = rtrie.vid_for_value(
vid16_nodes, vid16_kids, value)
else:
val = rtrie.vid_for_value(
vid_nodes, vid_kids, value)
elif column.is_lz4:
val = clz4.compress(value)
else:
val = value
# self.assertEqual(db.get(txn, i), val)
if ind_db is not None:
# row_id should be in bitmap too
if val in bitmaps:
bitmap = bitmaps[val]
else:
bitmap = BitSet()
bitmap.loads(ind_db.get(txn, val))
bitmaps[val] = bitmap
self.assertTrue(i in bitmap)
txn.commit()
env.close()
def test_marble_insert(self):
# test general infomation
self.assertEqual(self.n_inserted, len(_ALBUMS))
self.assertEqual(_NPARTITIONS, len(self.files))
part_id = {}
# test that each sub db is fine
for date, file in self.files.iteritems():
env, txn, dbs, meta = self.marble._open(file)
# check meta db
self.assertTrue(meta.contains(txn, "_vid_nodes"))
self.assertTrue(meta.contains(txn, "_vid_kids"))
self.assertTrue(meta.contains(txn, "_vid16_nodes"))
self.assertTrue(meta.contains(txn, "_vid16_kids"))
self.assertEqual(meta.get(txn, "name"), ujson.dumps("Collections"))
self.assertEqual(meta.get(txn, "partition"), ujson.dumps("date"))
self.assertEqual(meta.get(txn, "fields"), ujson.dumps(_FIELDS))
self.assertEqual(meta.get(txn, "_pdata"), ujson.dumps(date))
vid_nodes, _ = meta.get_raw(txn, '_vid_nodes')
vid_kids, _ = meta.get_raw(txn, '_vid_kids')
vid16_nodes, _ = meta.get_raw(txn, '_vid16_nodes', (None, 0))
vid16_kids, _ = meta.get_raw(txn, '_vid16_kids', (None, 0))
# check subdb, subinddb
part_id[date] = 1
for name, (db, ind_db, _, column, _) in dbs.iteritems():
if name == "_count":
continue
bitmaps = {}
part_id[date] = 1
for album in self.albums:
if date == album[_PARTITIONS]: # match the partition
value = album[name]
i = part_id[album[_PARTITIONS]]
part_id[album[_PARTITIONS]] += 1
if column.is_trie:
if column.rtrie_indicator == mdb.MDB_UINT_16:
val = rtrie.vid_for_value(vid16_nodes, vid16_kids, value)
else:
val = rtrie.vid_for_value(vid_nodes, vid_kids, value)
elif column.is_lz4:
val = clz4.compress(value)
else:
val = value
# self.assertEqual(db.get(txn, i), val)
if ind_db is not None:
# row_id should be in bitmap too
if val in bitmaps:
bitmap = bitmaps[val]
else:
bitmap = BitSet()
bitmap.loads(ind_db.get(txn, val))
bitmaps[val] = bitmap
self.assertTrue(i in bitmap)
txn.commit()
env.close()
def test_rtrie_in_mdb(self):
t = Trie()
self.assertEqual(t.add('hello'), 1)
self.assertEqual(t.add('hell'), 2)
self.assertEqual(t.add('hello'), 1)
self.assertEqual(t.add('hellothere'), 3)
self.assertEqual(t.add('good'), 4)
self.assertEqual(t.add('goodbye'), 5)
self.assertEqual(t.add('hello'), 1)
self.assertEqual(t.add('hellsink'), 6)
nodes, kids, _ = t.serialize()
nodeaddr, nodelen = nodes.buffer_info()
kidaddr, kidlen = kids.buffer_info()
try:
env = mdb.Env('/tmp/test_rtrie',
flags=mdb.MDB_WRITEMAP | mdb.MDB_NOSYNC
| mdb.MDB_NOSUBDIR)
txn = env.begin_txn()
db = env.open_db(txn, name='_meta_', flags=mdb.MDB_CREATE)
db.put_raw(txn, 'nodes', nodeaddr, nodelen)
db.put_raw(txn, 'kids', kidaddr, kidlen)
n, ns = db.get_raw(txn, 'nodes')
k, ks = db.get_raw(txn, 'kids')
txn.commit()
env.close()
env = mdb.Env('/tmp/test_rtrie',
flags=mdb.MDB_NOSYNC | mdb.MDB_NOSUBDIR)
txn = env.begin_txn()
db = env.open_db(txn, name='_meta_')
n, ns = db.get_raw(txn, 'nodes')
k, ks = db.get_raw(txn, 'kids')
self.assertEqual(rtrie.vid_for_value(n, k, 'hello'), 1)
self.assertEqual(rtrie.vid_for_value(n, k, 'hell'), 2)
self.assertEqual(rtrie.vid_for_value(n, k, 'goodbye'), 5)
self.assertEqual(rtrie.vid_for_value(n, k, 'hellsink'), 6)
self.assertEqual(rtrie.vid_for_value(n, k, 'hellothere'), 3)
self.assertEqual(rtrie.vid_for_value(n, k, 'good'), 4)
self.assertIsNone(rtrie.vid_for_value(n, k, 'notthere'))
txn.commit()
env.close()
finally:
import os
os.unlink('/tmp/test_rtrie')
os.unlink('/tmp/test_rtrie-lock')
def test_stress_wtrie(self):
ktrie = Trie()
strie = Trie()
etrie = Trie()
keywords = {}
search_terms = {}
exchange_ids = {}
with open(fixture) as f:
for data in f:
for word in data.split(' '):
vid = ktrie.add(word)
actual_vid = keywords.get(word)
if actual_vid is not None:
self.assertEqual(vid, actual_vid)
else:
keywords[word] = vid
vid = strie.add(data)
actual_vid = search_terms.get(data)
if actual_vid is not None:
self.assertEqual(vid, actual_vid)
else:
search_terms[data] = vid
nodes, kids, nodelen = etrie.serialize()
naddr, nlen = nodes.buffer_info()
kaddr, klen = kids.buffer_info()
#summarize(naddr, kaddr, nodelen)
#print_it(naddr, kaddr)
for dc, vid in exchange_ids.iteritems():
rvid = etrie.add(dc)
self.assertEqual(vid, rvid)
print dc, vid
value = value_for_vid(naddr, kaddr, vid)
self.assertEqual(dc, value)
if dc != value:
print " dc=%s adc=%s" % (dc, value)
avid = vid_for_value(naddr, kaddr, dc)
#print "vid=%s avid=%s" % (vid, avid)
self.assertEqual(vid, avid)
def test_stress_wtrie(self):
ktrie = Trie()
strie = Trie()
etrie = Trie()
keywords = {}
search_terms = {}
exchange_ids = {}
with open(fixture) as f:
for data in f:
for word in data.split(" "):
vid = ktrie.add(word)
actual_vid = keywords.get(word)
if actual_vid is not None:
self.assertEqual(vid, actual_vid)
else:
keywords[word] = vid
vid = strie.add(data)
actual_vid = search_terms.get(data)
if actual_vid is not None:
self.assertEqual(vid, actual_vid)
else:
search_terms[data] = vid
nodes, kids, nodelen = etrie.serialize()
naddr, nlen = nodes.buffer_info()
kaddr, klen = kids.buffer_info()
# summarize(naddr, kaddr, nodelen)
# print_it(naddr, kaddr)
for dc, vid in exchange_ids.iteritems():
rvid = etrie.add(dc)
self.assertEqual(vid, rvid)
print dc, vid
value = value_for_vid(naddr, kaddr, vid)
self.assertEqual(dc, value)
if dc != value:
print " dc=%s adc=%s" % (dc, value)
avid = vid_for_value(naddr, kaddr, dc)
# print "vid=%s avid=%s" % (vid, avid)
self.assertEqual(vid, avid)
def test_rtrie_in_mdb(self):
t = Trie()
self.assertEqual(t.add("hello"), 1)
self.assertEqual(t.add("hell"), 2)
self.assertEqual(t.add("hello"), 1)
self.assertEqual(t.add("hellothere"), 3)
self.assertEqual(t.add("good"), 4)
self.assertEqual(t.add("goodbye"), 5)
self.assertEqual(t.add("hello"), 1)
self.assertEqual(t.add("hellsink"), 6)
nodes, kids, _ = t.serialize()
nodeaddr, nodelen = nodes.buffer_info()
kidaddr, kidlen = kids.buffer_info()
try:
env = mdb.Env("/tmp/test_rtrie", flags=mdb.MDB_WRITEMAP | mdb.MDB_NOSYNC | mdb.MDB_NOSUBDIR)
txn = env.begin_txn()
db = env.open_db(txn, name="_meta_", flags=mdb.MDB_CREATE)
db.put_raw(txn, "nodes", nodeaddr, nodelen)
db.put_raw(txn, "kids", kidaddr, kidlen)
n, ns = db.get_raw(txn, "nodes")
k, ks = db.get_raw(txn, "kids")
txn.commit()
env.close()
env = mdb.Env("/tmp/test_rtrie", flags=mdb.MDB_NOSYNC | mdb.MDB_NOSUBDIR)
txn = env.begin_txn()
db = env.open_db(txn, name="_meta_")
n, ns = db.get_raw(txn, "nodes")
k, ks = db.get_raw(txn, "kids")
self.assertEqual(rtrie.vid_for_value(n, k, "hello"), 1)
self.assertEqual(rtrie.vid_for_value(n, k, "hell"), 2)
self.assertEqual(rtrie.vid_for_value(n, k, "goodbye"), 5)
self.assertEqual(rtrie.vid_for_value(n, k, "hellsink"), 6)
self.assertEqual(rtrie.vid_for_value(n, k, "hellothere"), 3)
self.assertEqual(rtrie.vid_for_value(n, k, "good"), 4)
self.assertIsNone(rtrie.vid_for_value(n, k, "notthere"))
txn.commit()
env.close()
finally:
import os
os.unlink("/tmp/test_rtrie")
os.unlink("/tmp/test_rtrie-lock")
def test_rtrie_in_memory(self):
s = unicode(u"séllsink").encode("utf-8")
# print "HELLSINK: %s" % s
t = Trie()
self.assertEqual(t.add("hello"), 1)
self.assertEqual(t.add("hell"), 2)
self.assertEqual(t.add("hello"), 1)
self.assertEqual(t.add("hellothere"), 3)
self.assertEqual(t.add("good"), 4)
self.assertEqual(t.add("goodbye"), 5)
self.assertEqual(t.add("hello"), 1)
self.assertEqual(t.add("hellsink"), 6)
self.assertEqual(t.add(s), 7)
t.print_it()
nodes, kids, _ = t.serialize()
nodeaddr, nodelen = nodes.buffer_info()
kidaddr, kidlen = kids.buffer_info()
print "LENS %s %s" % (nodelen, kidlen)
for i in range(8):
val = rtrie.value_for_vid(nodeaddr, kidaddr, i)
print "Value", i, val
self.assertEqual(rtrie.vid_for_value(nodeaddr, kidaddr, "hello"), 1)
self.assertEqual(rtrie.vid_for_value(nodeaddr, kidaddr, "hell"), 2)
self.assertEqual(rtrie.vid_for_value(nodeaddr, kidaddr, "goodbye"), 5)
self.assertEqual(rtrie.vid_for_value(nodeaddr, kidaddr, "hellsink"), 6)
self.assertEqual(rtrie.vid_for_value(nodeaddr, kidaddr, "hellothere"), 3)
self.assertEqual(rtrie.vid_for_value(nodeaddr, kidaddr, "good"), 4)
self.assertEqual(rtrie.vid_for_value(nodeaddr, kidaddr, s), 7)
self.assertIsNone(rtrie.vid_for_value(nodeaddr, kidaddr, "notthere"))
self.assertIsNone(rtrie.vid_for_value(nodeaddr, kidaddr, "h"))
self.assertIsNone(rtrie.vid_for_value(nodeaddr, kidaddr, "he"))
self.assertIsNone(rtrie.vid_for_value(nodeaddr, kidaddr, "hel"))
self.assertIsNone(rtrie.vid_for_value(nodeaddr, kidaddr, "hells"))
def test_rtrie_in_memory(self):
s = unicode(u'séllsink').encode('utf-8')
#print "HELLSINK: %s" % s
t = Trie()
self.assertEqual(t.add('hello'), 1)
self.assertEqual(t.add('hell'), 2)
self.assertEqual(t.add('hello'), 1)
self.assertEqual(t.add('hellothere'), 3)
self.assertEqual(t.add('good'), 4)
self.assertEqual(t.add('goodbye'), 5)
self.assertEqual(t.add('hello'), 1)
self.assertEqual(t.add('hellsink'), 6)
self.assertEqual(t.add(s), 7)
t.print_it()
nodes, kids, _ = t.serialize()
nodeaddr, nodelen = nodes.buffer_info()
kidaddr, kidlen = kids.buffer_info()
print "LENS %s %s" % (nodelen, kidlen)
for i in range(8):
val = rtrie.value_for_vid(nodeaddr, kidaddr, i)
print "Value", i, val
self.assertEqual(rtrie.vid_for_value(nodeaddr, kidaddr, 'hello'), 1)
self.assertEqual(rtrie.vid_for_value(nodeaddr, kidaddr, 'hell'), 2)
self.assertEqual(rtrie.vid_for_value(nodeaddr, kidaddr, 'goodbye'), 5)
self.assertEqual(rtrie.vid_for_value(nodeaddr, kidaddr, 'hellsink'), 6)
self.assertEqual(rtrie.vid_for_value(nodeaddr, kidaddr, 'hellothere'),
3)
self.assertEqual(rtrie.vid_for_value(nodeaddr, kidaddr, 'good'), 4)
self.assertEqual(rtrie.vid_for_value(nodeaddr, kidaddr, s), 7)
self.assertIsNone(rtrie.vid_for_value(nodeaddr, kidaddr, 'notthere'))
self.assertIsNone(rtrie.vid_for_value(nodeaddr, kidaddr, 'h'))
self.assertIsNone(rtrie.vid_for_value(nodeaddr, kidaddr, 'he'))
self.assertIsNone(rtrie.vid_for_value(nodeaddr, kidaddr, 'hel'))
self.assertIsNone(rtrie.vid_for_value(nodeaddr, kidaddr, 'hells'))