def prep_catalog():
"""Download python mailing list, create new catalog and catalog
messages, if not done already.
"""
if not os.path.exists(BENCHMARK_DATA_DIR):
os.makedirs(BENCHMARK_DATA_DIR)
# Check to see if mailing list data already present
if len(get_mailbox_filenames()) == 0:
MailListSucker(MAILLIST_INDEX, BENCHMARK_DATA_DIR).suck()
# Create ZODB and index maillist messages, if not yet done
zodb_file = os.path.join(BENCHMARK_DATA_DIR, 'test.zodb')
if not os.path.exists(zodb_file):
# Create a catalog
manager = ConnectionManager()
factory = FileStorageCatalogFactory(
os.path.join(BENCHMARK_DATA_DIR, 'test.zodb'), 'benchmark')
c = factory(manager)
# Create some indices
c['subject'] = CatalogFieldIndex(get_subject)
c['date'] = CatalogFieldIndex(get_date)
c['sender_email'] = CatalogFieldIndex(get_sender_email)
c['topics'] = CatalogFacetIndex(get_topics, topic_taxonomy)
c['text'] = CatalogTextIndex(get_text)
manager.commit()
# Loop over messages to get base line
profiler.start("Loop over messages without indexing")
for _ in MessageIterator():
pass
profiler.stop("Loop over messages without indexing")
profiler.start("Index messages")
id = 1
for msg in MessageIterator():
c.index_doc(id, msg)
id += 1
if id / 100 == 0:
manager.commit()
manager.commit()
manager.close()
profiler.stop("Index messages")
print("Indexed %d messages" % id)
def run():
# Download mailbox archive of python mailing list and build
# catalog if needed
prep_catalog()
# Open a catalog
manager = ConnectionManager()
factory = FileStorageCatalogFactory(
os.path.join(BENCHMARK_DATA_DIR, 'test.zodb'), 'benchmark')
c = factory(manager)
# Do some searches
profiler.start("unsorted retrieval")
n, results = c.search(date=('0', 'Z'))
print('%d results ' % n)
# Force generator to marshall brains
for result in results:
pass
profiler.stop("unsorted retrieval")
profiler.start("repeat unsorted retrieval")
n, results = c.search(date=('0', 'Z'))
print('%d results ' % n)
# Force generator to marshall brains
for result in results:
pass
profiler.stop("repeat unsorted retrieval")
profiler.start("sorted retrieval")
n, results = c.search(date=('0', 'Z'), sort_index='subject')
print('%d results ' % n)
for result in results:
pass
profiler.stop("sorted retrieval")
profiler.start("reverse sorted retrieval")
n, results = c.search(date=('0', 'Z'), sort_index='subject', reverse=True)
print('%d results ' % n)
for result in results:
pass
profiler.stop("reverse sorted retrieval")
profiler.start('limit to topic=year:2000')
n, results = c.search(topics=['year:2000'])
print('%d results' % n)
L = []
for result in results:
L.append(result)
profiler.stop("limit to topic=year:2000")
profiler.start('count limited to topic=year:2000')
print(c['topics'].counts(L, ['year:2000']))
profiler.stop('count limited to topic=year:2000')
profiler.stop()
profiler.print_stack()
def prep_catalog():
"""Download python mailing list, create new catalog and catalog
messages, if not done already.
"""
if not os.path.exists(BENCHMARK_DATA_DIR):
os.makedirs(BENCHMARK_DATA_DIR)
# Check to see if mailing list data already present
if len(get_mailbox_filenames()) == 0:
MailListSucker(MAILLIST_INDEX,BENCHMARK_DATA_DIR).suck()
# Create ZODB and index maillist messages, if not yet done
zodb_file = os.path.join(BENCHMARK_DATA_DIR, 'test.zodb')
if not os.path.exists(zodb_file):
# Create a catalog
manager = ConnectionManager()
factory = FileStorageCatalogFactory(
os.path.join(BENCHMARK_DATA_DIR,
'test.zodb'), 'benchmark' )
c = factory(manager)
# Create some indices
c['subject'] = CatalogFieldIndex(get_subject)
c['date'] = CatalogFieldIndex(get_date)
c['sender_email'] = CatalogFieldIndex(get_sender_email)
c['topics'] = CatalogFacetIndex(get_topics, topic_taxonomy)
c['text'] = CatalogTextIndex(get_text)
manager.commit()
# Loop over messages to get base line
profiler.start( "Loop over messages without indexing" )
for _ in MessageIterator():
pass
profiler.stop( "Loop over messages without indexing" )
profiler.start( "Index messages" )
id = 1
for msg in MessageIterator():
c.index_doc(id,msg)
id += 1
if id / 100 == 0:
manager.commit()
manager.commit()
manager.close()
profiler.stop( "Index messages" )
print "Indexed %d messages" % id
def do_benchmark(fname, nd, nk1, nk2, out=sys.stdout):
cumulative1 = 0.0
cumulative2 = 0.0
print("Index 1:", file=out)
print("\t# docs: %d" % nd, file=out)
print("\t# distinct keys: %d" % nk1, file=out)
print("Index 2:", file=out)
print("\t# docs: %d" % nd, file=out)
print("\t# distinct keys: %d" % nk2, file=out)
print("", file=out)
cost1, cost2 = predictions(nd, nk1, nk2)
print('Cost1: %0.2f' % cost1, file=out)
print('Cost2: %0.2f' % cost2, file=out)
print("Prediction:", file=out)
if cost1 > cost2:
print("Algorithm 2 %0.2f times faster than Algorithm 1" %
(cost1 / cost2),
file=out)
else:
print("Algorithm 1 %0.2f times faster than Algorithm 2" %
(cost2 / cost1),
file=out)
print("", file=out)
print("Setting up indexes...", file=out)
for fn in glob.glob(fname + "*"):
os.remove(fn)
manager = ConnectionManager()
factory = FileStorageCatalogFactory(fname, 'intersection')
catalog = factory(manager)
catalog['one'] = CatalogFieldIndex('one')
catalog['two'] = CatalogFieldIndex('two')
class Document(object):
def __init__(self, docid):
self.one = str(docid % nk1)
self.two = str(docid % nk2)
for docid in xrange(nd):
catalog.index_doc(docid, Document(docid))
manager.commit()
manager.close()
N_QUERIES = 1000
print("Running %d queries for each algorithm..." % N_QUERIES, file=out)
catalog = factory(manager)
for _ in xrange(1000):
key1 = random.randrange(nk1)
key2 = random.randrange(nk2)
query1 = Intersection1(Eq('one', str(key1)), Eq('two', str(key2)))
query2 = Intersection2(Eq('one', str(key1)), Eq('two', str(key2)))
start = time.time()
result1 = query1.apply(catalog)
cumulative1 += time.time() - start
start = time.time()
result2 = query2.apply(catalog)
cumulative2 += time.time() - start
s1 = sorted(list(result1))
s2 = sorted(list(result2))
assert s1 == s2, (s1, s2)
manager.close()
for fn in glob.glob(fname + "*"):
os.remove(fn)
print("", file=out)
print("Result:", file=out)
print("Time for algorithm1: %0.3f s" % cumulative1, file=out)
print("Time for algorithm2: %0.3f s" % cumulative2, file=out)
if cumulative1 > cumulative2:
print("Algorithm 2 %0.2f times faster than Algorithm 1" %
(cumulative1 / cumulative2),
file=out)
else:
print("Algorithm 1 %0.2f times faster than Algorithm 2" %
(cumulative2 / cumulative1),
file=out)
return cost1 / cost2, cumulative1 / cumulative2
def do_benchmark(fname, nd, nk1, nk2, out=sys.stdout):
cumulative1 = 0.0
cumulative2 = 0.0
print >>out, "Index 1:"
print >>out, "\t# docs: %d" % nd
print >>out, "\t# distinct keys: %d" % nk1
print >>out, "Index 2:"
print >>out, "\t# docs: %d" % nd
print >>out, "\t# distinct keys: %d" % nk2
print >>out, ""
cost1, cost2 = predictions(nd, nk1, nk2)
print >>out, 'Cost1: %0.2f' % cost1
print >>out, 'Cost2: %0.2f' % cost2
print >>out
print >>out, "Prediction:"
if cost1 > cost2:
print >>out, "Algorithm 2 %0.2f times faster than Algorithm 1" % (
cost1/cost2)
else:
print >>out, "Algorithm 1 %0.2f times faster than Algorithm 2" % (
cost2/cost1)
print >>out, ""
print >>out, "Setting up indexes..."
for fn in glob.glob(fname + "*"):
os.remove(fn)
manager = ConnectionManager()
factory = FileStorageCatalogFactory(fname, 'intersection')
catalog = factory(manager)
catalog['one'] = CatalogFieldIndex('one')
catalog['two'] = CatalogFieldIndex('two')
class Document(object):
def __init__(self, docid):
self.one = str(docid % nk1)
self.two = str(docid % nk2)
for docid in xrange(nd):
catalog.index_doc(docid, Document(docid))
manager.commit()
manager.close()
N_QUERIES = 1000
print >>out, "Running %d queries for each algorithm..." % N_QUERIES
catalog = factory(manager)
for _ in xrange(1000):
key1 = random.randrange(nk1)
key2 = random.randrange(nk2)
query1 = Intersection1(Eq('one', str(key1)), Eq('two', str(key2)))
query2 = Intersection2(Eq('one', str(key1)), Eq('two', str(key2)))
start = time.time()
result1 = query1.apply(catalog)
cumulative1 += time.time() - start
start = time.time()
result2 = query2.apply(catalog)
cumulative2 += time.time() - start
s1 = sorted(list(result1))
s2 = sorted(list(result2))
assert s1==s2, (s1, s2)
manager.close()
for fn in glob.glob(fname + "*"):
os.remove(fn)
print >>out, ""
print >>out, "Result:"
print >>out, "Time for algorithm1: %0.3f s" % cumulative1
print >>out, "Time for algorithm2: %0.3f s" % cumulative2
if cumulative1 > cumulative2:
print >>out, "Algorithm 2 %0.2f times faster than Algorithm 1" % (
cumulative1/cumulative2)
else:
print >>out, "Algorithm 1 %0.2f times faster than Algorithm 2" % (
cumulative2/cumulative1)
return cost1 / cost2, cumulative1 / cumulative2