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
|
#!/usr/bin/env python
# Benchmark the times reading large datasets with Blosc and Blosc2 filters.
import sys
from time import time
import numpy as np
import pandas as pd
import tables as tb
def time_inkernel(table_blosc):
t1 = time()
res = [
x["precip"]
for x in table_blosc.where(
"(lat > 50) & (20 <= lon) & (lon < 50) & (time < 10)"
)
]
# res = [x['precip'] for x in table_blosc.where("(time < 10)")]
print(len(res))
return time() - t1
def time_read(table):
n_reads = 10_000
t0 = time()
idxs_to_read = np.random.randint(0, table.nrows, n_reads)
# print(f"Time to create indexes: {time() - t0:.3f}s")
print(f"Randomly reading {n_reads // 1_000} Krows...", end="")
t0 = time()
for i in idxs_to_read:
_ = table[i]
t = time() - t0
print(f"\t{t:.3f}s ({t / n_reads * 1e6:.1f} us/read)")
print(f"Querying {table.nrows // 1000_000_000} Grows...", end="")
t0 = time()
_ = [x["precip"] for x in table.where("(time < 10)")]
t = time() - t0
print(
f"\t\t{t:.3f}s ({table.nrows * table.dtype.itemsize / t / 2**30:.1f} GB/s)"
)
def time_pandas(df):
t1 = time()
res = df.query("(lat > 50) & (20 <= lon) & (lon < 50) & (time < 10)")[
"precip"
]
print(len(res))
return time() - t1
def pandas_create_df():
f = tb.open_file("wblosc_table.h5", "r")
df = pd.DataFrame(f.root.table_blosc[:])
f.close()
return df
def inkernel_blosc2_blosclz(table):
print(
f"Time to read 6 inkernel queries with Blosc2 (blosclz): {time_inkernel(table):.3f} sec"
)
def inkernel_blosc2_lz4(table):
print(
f"Time to read 6 inkernel queries with Blosc2 (lz4): {time_inkernel(table):.3f} sec"
)
def pandas_query_numexpr(df):
print(
f"Time to perform 6 pandas+numexpr queries: {time_pandas(df):.3f} sec"
)
f = tb.open_file(sys.argv[1])
table = f.root.table_blosc
time_read(table)
f.close()
# df = pandas_create_df()
# pandas_query_numexpr(df)
|
