-
Notifications
You must be signed in to change notification settings - Fork 1
/
benchmark.py
185 lines (158 loc) · 10.4 KB
/
benchmark.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
import json
import pandas as pd
import mysql.connector
from datetime import datetime
from weaviate_import import init_client
from timeit import default_timer as timer
from utils import add_limit, build_sql_query, get_operands, get_track_names, save_results
# Example: Queen - Get Down, Make Love - Remastered 2011
example_vector = [-0.06430846, -0.24293041, 0.110439375, -0.100999914, 0.09629649, -0.04072969, -0.169328, -0.014131491, 0.117915705, 0.5358003, 0.13591173, 0.1179824, -0.08689239, 0.060807075, 0.09822265, 0.19681141, -0.14886066, 0.060191516, -0.46446335, -0.016015284, 0.06305353, 0.14932896, -0.25746924, 0.1558214, -0.2350968, -0.44718155, 0.27163702, -0.54405105, -0.09141091, 0.050858628, 0.34632823, 0.48786363, 0.32426932, 0.5739468, 0.23869427, -0.3895692, 0.04086138, 0.30526224, -0.09721386, 0.7694963, -0.23109816, 0.25778306, 0.06246513, -0.15824226, -0.07554409, 0.4262448, 0.3915728, -0.10568536, -0.3647673, 0.012407957, -0.6349575, -0.2149216, -0.2833733, -0.45024365, -0.09075926, 0.13799985, -0.096156955, -0.15037905, 0.1893991, 0.1821173, 0.57067454, -0.8982235, -0.4192168, -0.25409874, 0.298224, -0.22855514, -0.17255244, -0.2956281, -0.19140317, -0.08400424, -0.14878975, 0.08089396, 0.11062146, -0.19968952, 0.5094097, -0.5838567, -0.17085947, 0.096286826, 0.2912769, 0.11041665, -0.09463574, -0.781021, 0.15997428, 0.11306844, 0.78838485, 0.046952806, 0.20534825, -0.013530162, 0.13570505, 0.39102194, -0.01179285, 0.21663925, -0.24078456, 0.30120635, -0.027091846, 0.14774093, -0.29773268, -0.42512724, -0.11435961, 0.18203783, -0.41414407, -0.20505615, 0.8132204, 0.018949207, -0.14595698, -0.15278365, -0.112690575, -0.5838136, -0.21263546, -0.13692501, -0.008073677, -0.13169445, 0.24559294, 0.11468319, -0.30848005, -0.56463796, 0.080196455, -0.13245742, -0.1404417, 0.081029356, 0.18065631, 0.31255183, 0.10336783, 0.13753967, 0.022072855, 0.14859243, 1.6029545, 0.46531165, -0.20865615, 0.012292662, 0.15616858, -0.45227888, -0.77387786, -0.43062842, 0.016371612, 0.3711838, 0.5039181, 0.15240684, -0.1913129, -0.27780148, -0.009997403, 0.117378555, -0.0085565215, -0.07294885, -0.024712797, 0.42950535, -0.005417766, -0.041284315, -0.1248916, -0.09128173, 0.345162, 0.2769205, -0.3914345, 0.057552766, -0.027560443, 0.5258812, 0.10822242, 0.31862304, -0.6082464, -0.22113307, -0.6233102, -0.34389177, -0.1452825, -0.1252571, -0.12338128, 0.4402884, 0.34804556, -0.13827753, -0.08136732, -0.306151, -0.18906707, 0.39232698, -0.09769015, 0.18603276, -0.517225, -0.7027275, 0.034595076, -0.24334815, 0.17681517, 0.28136656, -0.35619092, -1.133969, 0.057308406, -0.5783666, 0.10168155, -0.109514184, -0.23636238, 0.091893405, 0.086915925, 0.54031307, 0.13154425, 0.39741704, 0.17699283, 0.339236, -0.2782195, 0.0015203763, -0.29165968, 0.35996062, -0.083890885, -0.030090114, -0.43646154, -0.17463323, 0.40066385, -0.087886214, -0.18553457, 0.39516634, 0.26139808, -0.31970882, 0.35820448, 0.10034503, -0.3061243, -0.036659196, -0.10865491, 0.4053416, -0.11471894, 0.43889588, 0.36892733, -0.12887777, 0.055665467, -0.047561377, -0.109102905, -0.08950873, -0.057528242, -0.2530506, 0.1425888, -0.11001377, -0.5834436, 0.20523982, -0.23291121, -0.41186935, -0.23352517, -0.41898695, 0.24615884, 0.63720244, 0.08238311, -0.32418433, 0.19017169, 0.3589935, 0.12049238, 0.33731705, -0.01893071, -0.1797846, 0.4786759, 0.20034383, 0.18493232, 0.09827406, 0.14285572, -0.17101601, 0.46751156, 0.12339927, -0.076471925, 0.2264648, 0.43471265, 0.10593008, 0.2955751, -0.08592804, -0.2486908, 0.091301665, -0.5648791, 0.036137637, 0.025429724, 0.2128401, 0.35821828, -0.2509771, -0.26594982, 0.18869537, -0.21725583, -0.35601154, -0.36929485, -0.29925507, -0.47262815, -0.041473743, 0.10217856, 0.062293913, -0.15322512, 0.08956682, -0.35481754, 0.058380894, 0.5370905, 0.40207297, 0.10658187, -0.31264502, 0.22212799, -0.05873761, 0.28276542, 0.04365597, 0.23748645, 0.29086804, 0.10375105, 0.32179505, -0.10878695, 0.17493275, 0.10128974, -0.046946246, -0.33823404, 0.33285227, -0.054822538, 0.25945452, 0.07178426, -0.14386761]
sql_columns = ["track_name", "track_artist", "track_album_name", "lyrics", "playlist_genre", "playlist_subgenre"]
columns = ["track_name", "artist_name", "album_name", "lyrics", "genre", "subgenre"]
limits = [10, 20, 30]
databases = ["document", "sql_single_table", "sql", "vector"]
benchmarks = {
"BM1": { # Benchmark 1: Search for a single word in a single column
"query": "love",
"is_array": False,
"columns": columns
},
"BM2": { # Benchmark 2: Search for a single word in all columns
"query": ["Love", "Cake", "Fame"],
"is_array": True,
"columns": columns
},
"BM3": { # Benchmark 3: Search for sentence in all columns
"query": "What is love",
"is_array": False,
"columns": columns
},
"BM4": { # Benchmark 4: Search in single column
"query": "love",
"is_array": False,
"columns": ["track_name"]
},
"BM5": { # Benchmark 5: Search for song with vector
"query": example_vector,
"is_array": False,
"columns": columns
}
}
def aggregate_query_time(func, n=20):
'''
Aggregate the query time for a function
'''
total_time = 0
for i in range(n):
time = measure(func)
total_time += time
# print (f"Query {i}/{n} took {time} seconds")
return total_time / n
def measure(func):
'''
Measure execution time of a function
'''
start = timer()
func()
end = timer()
return end - start
def executeSql(cursor, query):
cursor.execute(query)
return cursor.fetchall()
def run_benchmark(query, benchmark_id, database, limit):
query_time = aggregate_query_time(lambda: query.do())
result = query.do()
count = len(result["data"]["Get"]["Track"])
result = get_track_names(sorted([t["track_name"] for t in result["data"]["Get"]["Track"]])[:10])
df = pd.DataFrame({"id": benchmark_id, "database": database, "avg_query_time": query_time, "count_result": count, "result": result, "limit": limit}, index=[0])
print(f"- Time for {database} database: {query_time}")
return df
def run_benchmark_sql(q, benchmark_id, database, limit):
try:
cnx = mysql.connector.connect(user='root', password='root', host='127.0.0.1', database='spotifyDataset')
cursor = cnx.cursor()
query_time = aggregate_query_time(func=(lambda: executeSql(cursor, query = q)))
result = executeSql(cursor, q)
cnx.close()
count = len(result)
result = get_track_names(sorted(t[0] for t in result)[:10])
df = pd.DataFrame({"id": benchmark_id, "database": database, "avg_query_time": query_time, "count_result": count, "result": result , "limit": limit}, index=[0])
print(f"- Time for {database} database: {query_time}")
return df
except mysql.connector.Error as err:
print("Something went wrong: {}".format(err))
def main(args):
'''
Benchmark weaviate vs. SQL similiarity query performance
Query the databases and measure the execution time
for Vector, Document and SQL databases.
Run benchmark 100x and take the average
We evaluate the quality of results ourselves
'''
# init dbs
weaviate_client = init_client()
df = pd.DataFrame(columns=["id", "database", "avg_query_time", "count_result","result", "limit"])
# query benchmarks
for key, benchmark in benchmarks.items():
query = benchmark["query"]
is_array = benchmark["is_array"]
columns = benchmark["columns"]
for limit in limits:
print("-----------------------------------------------------------")
print(f"{key} with limit {limit}")
print("-----------------------------------------------------------")
for database in databases:
print(f"Measuring benckmark for {database} database ...")
if database == "vector":
if key == 'BM5':
q = (
weaviate_client.query
.get("Track", ["track_name"])
.with_near_vector({ "vector": example_vector})
)
else:
nearText = {
"concepts": query if is_array else [query],
}
q = (
weaviate_client.query
.get("Track", ["track_name"])
.with_near_text(nearText)
)
if limit > 1:
q = q.with_limit(limit)
result = run_benchmark(query=q, benchmark_id=key, database=database, limit=limit)
elif database == "document":
if key == 'BM5':
continue
withFilter = {
"operator": "Or",
"operands": get_operands(query, columns=columns, is_array=is_array)
}
q = (
weaviate_client.query
.get("Track", ["track_name"])
.with_where(withFilter)
)
if limit > 1:
q = q.with_limit(limit)
result = run_benchmark(query=q, benchmark_id=key, database=database, limit=limit)
elif database == "sql":
if key == 'BM5':
continue
q = build_sql_query(query, columns = (columns if key=='BM4' else sql_columns), limit=limit, is_array=is_array)
result = run_benchmark_sql(q, benchmark_id=key, database=database, limit=limit)
elif database == "sql_single_table":
if key == 'BM5':
continue
q = build_sql_query(query, columns = (columns if key=='BM4' else sql_columns), limit=limit, is_array=is_array, is_singelTable=True)
result = run_benchmark_sql(q, benchmark_id=key, database=database, limit=limit)
df = pd.concat([df, result], ignore_index=True)
print("-----------------------------------------------------------")
save_results(df)
if __name__ == "__main__":
import argparse
parser = argparse.ArgumentParser(
description="Benchmark weaviate vs. SQL vs. Document query performance")
args = parser.parse_args()
main(args)