Python serialization benchmark

bench.py

"""
Python serialization benchmark
Based on https://gist.github.com/marians/f1314446b8bf4d34e782
2014-11-11 v1.0
2014-11-12 v1.1 Added compression and output size.
"""
try: import cPickle
except: import pickle as cPickle  # don't break in Python 3
import json, marshal, pickle, random
from hashlib import md5
from time import time
from zlib import compress, decompress  # halves json float size (19 kB to 10 kB)

#import ujson  # from http://www.lfd.uci.edu/~gohlke/pythonlibs/#ujson
# 2 to 3 times slower and up to 50% larger output than marshal.

data_size = 100000
test_runs = 10

if __name__ == "__main__":
	payload = {
		"float": [(random.randrange(0, 99) + random.random()) for i in range(data_size)],
		"int": [random.randrange(0, 9999) for i in range(data_size)],
		"str": [md5(str(random.random()).encode('utf8')).hexdigest() for i in range(data_size)]
	}
	modules = [json, pickle, cPickle, marshal]
	print("%s times %s elements:" % (test_runs, data_size))
	for payload_type in payload:
		data = payload[payload_type]
		for module in modules:
			dumps_binary = module.__name__ in ['pickle', 'cPickle', 'marshal']
			dumps = module.dumps
			loads = module.loads
			start = time()
			if module.__name__ in ['pickle', 'cPickle']:
				for i in range(test_runs): serialized = compress(dumps(data, protocol=-1))  # -1 = highest available (2 in Python 2; 3 in Python 3.4 (2x speed))
			elif dumps_binary:
				#print("%s kB" % (len(module.dumps(data))/1000))
				for i in range(test_runs): serialized = compress(dumps(data))
			else:
				for i in range(test_runs): serialized = compress(dumps(data).encode('utf8'))
			w = time() - start
			start = time()
			if dumps_binary:
				for i in range(test_runs): unserialized = loads(decompress(serialized))
			else:
				for i in range(test_runs): unserialized = loads(decompress(serialized).decode('utf8'))
			r = time() - start
			print("%s\t%s\t%6.3f sW %6.3f sR %10.0f kB" % (module.__name__, payload_type, w, r, len(serialized)/1e3))

result.txt

32-bit results in 64-bit Windows 7 Enterprise on Xeon W3540 @ 2.93 GHz 2.93 GHz (Dell Precision T3500):

C:\Python27\python.exe -u "serialization_benchmark.py"
1000 times 1000 elements:
json int W 0.125 R 0.156
pickle int W 2.808 R 1.139
cPickle int W 0.047 R 0.046
marshal int W 0.016 R 0.031
json float W 1.981 R 0.624
pickle float W 2.607 R 1.092
cPickle float W 0.063 R 0.062
marshal float W 0.047 R 0.031
json str W 0.172 R 0.437
pickle str W 5.149 R 2.309
cPickle str W 0.281 R 0.156
marshal str W 0.109 R 0.047

C:\pypy-1.6\pypy-c -u "serialization_benchmark.py"
1000 times 1000 elements:
json int W 0.515 R 0.452
pickle int W 0.546 R 0.219
cPickle int W 0.577 R 0.171
marshal int W 0.032 R 0.031
json float W 2.390 R 1.341
pickle float W 0.656 R 0.436
cPickle float W 0.593 R 0.406
marshal float W 0.327 R 0.203
json str W 1.141 R 1.186
pickle str W 0.702 R 0.546
cPickle str W 0.828 R 0.562
marshal str W 0.265 R 0.078

c:\Python34\python -u "serialization_benchmark.py"
1000 times 1000 elements:
json int	W 0.203 R 0.140		 6 kB
pickle int	W 0.047 R 0.062		 3 kB
pickle int	W 0.031 R 0.062		 3 kB
marshal int	W 0.031 R 0.047		 5 kB
json float	W 1.935 R 0.749		19 kB
pickle float	W 0.047 R 0.062		 9 kB
pickle float	W 0.047 R 0.062		 9 kB
marshal float	W 0.047 R 0.047		 9 kB
json str	W 0.281 R 0.187		36 kB
pickle str	W 0.125 R 0.140		35 kB
pickle str	W 0.125 R 0.140		35 kB
marshal str	W 0.094 R 0.078		34 kB

v1.1 zlib:
c:\Python34\python -u "serialization_benchmark.py"
1000 times 1000 elements:
json	int	 0.686 sW  0.187 sR          3 kB
pickle	int	 0.187 sW  0.078 sR          2 kB
pickle	int	 0.187 sW  0.094 sR          2 kB
marshal	int	 0.702 sW  0.094 sR          3 kB
json	float	 3.325 sW  0.874 sR         10 kB
pickle	float	 0.546 sW  0.140 sR          8 kB
pickle	float	 0.546 sW  0.125 sR          8 kB
marshal	float	 0.468 sW  0.109 sR          8 kB
json	str	 2.590 sW  0.608 sR         20 kB
pickle	str	 2.496 sW  0.452 sR         20 kB
pickle	str	 2.418 sW  0.421 sR         20 kB
marshal	str	 1.981 sW  0.374 sR         20 kB

v1.1 bz2:
c:\Python34\python -u "serialization_benchmark.py"
1000 times 1000 elements:
json	int	 1.188 sW  0.446 sR          2 kB
pickle	int	 1.263 sW  0.317 sR          2 kB
pickle	int	 1.295 sW  0.271 sR          2 kB
marshal	int	 1.627 sW  0.270 sR          2 kB
json	float	 4.828 sW  1.706 sR          8 kB
pickle	float	 2.737 sW  0.749 sR          8 kB
pickle	float	 2.699 sW  0.718 sR          8 kB
marshal	float	 2.590 sW  0.733 sR          8 kB
json	str	 5.179 sW  1.778 sR         17 kB
pickle	str	 4.883 sW  1.888 sR         17 kB
pickle	str	 4.849 sW  1.857 sR         17 kB
marshal	str	 4.782 sW  1.741 sR         17 kB

c:\Python34\python -u "serialization_benchmark_nocompress.py"
10 times 100000 elements:
json	int	 0.187 sW  0.140 sR        589 kB
pickle	int	 0.031 sW  0.062 sR        298 kB
pickle	int	 0.031 sW  0.047 sR        298 kB
marshal	int	 0.047 sW  0.047 sR        500 kB
json	float	 1.778 sW  0.686 sR       1919 kB
pickle	float	 0.062 sW  0.047 sR        900 kB
pickle	float	 0.047 sW  0.062 sR        900 kB
marshal	float	 0.047 sW  0.031 sR        900 kB
json	str	 0.312 sW  0.203 sR       3600 kB
pickle	str	 0.250 sW  0.156 sR       3501 kB
pickle	str	 0.250 sW  0.156 sR       3501 kB
marshal	str	 0.125 sW  0.094 sR       3400 kB

2.7 zlib:
C:\Python27\python.exe -u "serialization_benchmark.py"
10 times 100000 elements:
json	int	 1.108 sW  0.814 sR        226 kB
pickle	int	 3.011 sW  1.186 sR        216 kB
cPickle	int	 0.265 sW  0.094 sR        216 kB
marshal	int	 1.154 sW  0.078 sR        239 kB
json	float	 4.509 sW  1.030 sR        906 kB
pickle	float	 4.025 sW  1.185 sR        797 kB
cPickle	float	 1.498 sW  0.156 sR        797 kB
marshal	float	 1.108 sW  0.125 sR        793 kB
json	str	 3.276 sW  0.780 sR       1982 kB
pickle	str	 9.034 sW  2.981 sR       2259 kB
cPickle	str	 4.135 sW  0.686 sR       2259 kB
marshal	str	 2.980 sW  0.421 sR       1989 kB

zlib:
c:\Python34\python -u "serialization_benchmark.py"
10 times 100000 elements:
json	int	 1.123 sW  0.218 sR        226 kB
pickle	int	 0.265 sW  0.094 sR        216 kB
pickle	int	 0.281 sW  0.078 sR        216 kB
marshal	int	 1.170 sW  0.078 sR        242 kB
json	float	 4.227 sW  0.842 sR        906 kB
pickle	float	 1.529 sW  0.125 sR        797 kB
pickle	float	 1.529 sW  0.125 sR        797 kB
marshal	float	 1.154 sW  0.109 sR        794 kB
json	str	 3.182 sW  0.468 sR       1983 kB
pickle	str	 3.026 sW  0.452 sR       1966 kB
pickle	str	 3.074 sW  0.437 sR       1966 kB
marshal	str	 2.637 sW  0.343 sR       1933 kB

bz2:
c:\Python34\python -u "serialization_benchmark.py"
10 times 100000 elements:
json	int	 1.061 sW  0.655 sR        177 kB
pickle	int	 0.499 sW  0.218 sR        171 kB
pickle	int	 0.499 sW  0.234 sR        171 kB
marshal	int	 0.608 sW  0.374 sR        169 kB
json	float	 4.415 sW  2.683 sR        769 kB
pickle	float	 1.981 sW  1.170 sR        758 kB
pickle	float	 1.622 sW  0.920 sR        758 kB
marshal	float	 1.982 sW  1.185 sR        770 kB
json	str	 5.647 sW  3.931 sR       1711 kB
pickle	str	 5.381 sW  3.759 sR       1712 kB
pickle	str	 5.335 sW  3.806 sR       1712 kB
marshal	str	 5.179 sW  3.650 sR       1711 kB