Python 2.7.2 serializer speed comparisons

Jul 25, 2012

In a recent python project where I was sending multiple messages per second of data over a basic socket, I had initially just grabbed the cPickle module to get the prototype proof-of-concept functioning properly. cPickle is awesome for easily serializing more complex python objects like custom classes, even though in my case I am only sending basic types.

My messages were dicts with some nested dicts, lists, floats, and string values. Roughly 500-1000 bytes. cPickle was doing just fine, but there came a point where I wanted to investigate the areas that could be tightened up. The first thing I realized was that I had forgotten to encode cPickle in the binary format (the default is ascii). That saved me quite a bit of time. But then I casually searched online to see if any json options might be better since my data is pretty primitive anyways.

I found UltraJSON, which is a pure C json parsing library for python, and ran some tests. There are benchmarks on the project page for ujson, as well as other articles on the internet, but I just wanted to post up my own results using a mixed type data container. ujson came out extremely fast: faster than binary cPickle and msgpack, in the encoding test. Although in the decoding test, msgpack appeared to be fastest, followed by binary cPickle, and then ujson coming in 3rd

This test included the following:

Here is my Python 2.7.2 test script using timeit for each encode and decode step.

Dependencies:

pip install tabulate simplejson python-cjson ujson yajl msgpack-python

Test:

from timeit import timeit
from tabulate import tabulate

setup = '''
  d = {
    'words': """
      Lorem ipsum dolor sit amet, consectetur adipiscing
      elit. Mauris adipiscing adipiscing placerat.
      Vestibulum augue augue,
      pellentesque quis sollicitudin id, adipiscing.
      """,
    'list': range(100),
    'dict': dict((str(i),'a') for i in xrange(100)),
    'int': 100,
    'float': 100.123456
  }
'''

setup_pickle    = '%s ; import cPickle ; src = cPickle.dumps(d)' % setup
setup_pickle2   = '%s ; import cPickle ; src = cPickle.dumps(d, 2)' % setup
setup_json      = '%s ; import json; src = json.dumps(d)' % setup
setup_msgpack   = '%s ; src = msgpack.dumps(d)' % setup

tests = [
  # (title, setup, enc_test, dec_test)
  ('pickle (ascii)', 'import pickle; %s' % setup_pickle, 'pickle.dumps(d, 0)', 'pickle.loads(src)'),
  ('pickle (binary)', 'import pickle; %s' % setup_pickle2, 'pickle.dumps(d, 2)', 'pickle.loads(src)'),
  ('cPickle (ascii)', 'import cPickle; %s' % setup_pickle, 'cPickle.dumps(d, 0)', 'cPickle.loads(src)'),
  ('cPickle (binary)', 'import cPickle; %s' % setup_pickle2, 'cPickle.dumps(d, 2)', 'cPickle.loads(src)'),
  ('json', 'import json; %s' % setup_json, 'json.dumps(d)', 'json.loads(src)'),
  ('simplejson-3.3.1', 'import simplejson; %s' % setup_json, 'simplejson.dumps(d)', 'simplejson.loads(src)'),
  ('python-cjson-1.0.5', 'import cjson; %s' % setup_json, 'cjson.encode(d)', 'cjson.decode(src)'),
  ('ujson-1.33', 'import ujson; %s' % setup_json, 'ujson.dumps(d)', 'ujson.loads(src)'),
  ('yajl 0.3.5', 'import yajl; %s' % setup_json, 'yajl.dumps(d)', 'yajl.loads(src)'),
  ('msgpack-python-0.3.0', 'import msgpack; %s' % setup_msgpack, 'msgpack.dumps(d)', 'msgpack.loads(src)'),
]

loops = 15000
enc_table = []
dec_table = []

print "Running tests (%d loops each)" % loops

for title, mod, enc, dec in tests:
    print title

    print "  [Encode]", enc 
    result = timeit(enc, mod, number=loops)
    enc_table.append([title, result])

    print "  [Decode]", dec 
    result = timeit(dec, mod, number=loops)
    dec_table.append([title, result])

enc_table.sort(key=lambda x: x[1])
enc_table.insert(0, ['Package', 'Seconds'])

dec_table.sort(key=lambda x: x[1])
dec_table.insert(0, ['Package', 'Seconds'])

print "\nEncoding Test (%d loops)" % loops
print tabulate(enc_table, headers="firstrow")

print "\nDecoding Test (%d loops)" % loops
print tabulate(dec_table, headers="firstrow")

OUTPUT:

Running tests (15000 loops each)
pickle (ascii)
  [Encode] pickle.dumps(d, 0)
  [Decode] pickle.loads(src)
pickle (binary)
  [Encode] pickle.dumps(d, 2)
  [Decode] pickle.loads(src)
cPickle (ascii)
  [Encode] cPickle.dumps(d, 0)
  [Decode] cPickle.loads(src)
cPickle (binary)
  [Encode] cPickle.dumps(d, 2)
  [Decode] cPickle.loads(src)
json
  [Encode] json.dumps(d)
  [Decode] json.loads(src)
simplejson-3.3.1
  [Encode] simplejson.dumps(d)
  [Decode] simplejson.loads(src)
python-cjson-1.0.5
  [Encode] cjson.encode(d)
  [Decode] cjson.decode(src)
ujson-1.33
  [Encode] ujson.dumps(d)
  [Decode] ujson.loads(src)
yajl 0.3.5
  [Encode] yajl.dumps(d)
  [Decode] yajl.loads(src)
msgpack-python-0.3.0
  [Encode] msgpack.dumps(d)
  [Decode] msgpack.loads(src)

Encoding Test (15000 loops)
Package                 Seconds
--------------------  ---------
ujson-1.33             0.232215
msgpack-python-0.3.0   0.241945
cPickle (binary)       0.305273
yajl 0.3.5             0.634148
python-cjson-1.0.5     0.680604
json                   0.780438
simplejson-3.3.1       1.04763
cPickle (ascii)        1.62062
pickle (ascii)        14.0497
pickle (binary)       15.4712

Decoding Test (15000 loops)
Package                 Seconds
--------------------  ---------
msgpack-python-0.3.0   0.240885
cPickle (binary)       0.393152
ujson-1.33             0.396875
python-cjson-1.0.5     0.694321
yajl 0.3.5             0.748369
simplejson-3.3.1       0.780531
cPickle (ascii)        1.38561
json                   1.65921
pickle (binary)        5.20554
pickle (ascii)        17.8767

JustinFX.com

Python 2.7.2 serializer speed comparisons