I needed a package that decodes JSON into dictionaries/lists etc.  I’ve used simplejson in the past and it’s worked fairly well.  I came across another package today that claims to be at least a hundred times faster than the other JSON implementations for python called cjson.  I was a bit skeptical because simplejson has C extensions too.

I decided to run some quick benchmarks.  The code I ran can be found here.

benchmark data:
processor: Intel(R) Core(TM)2 CPU 6600 @ 2.40GHz
x86_64 Linux.
simplejson 1.7.3 (I assume the c extensions are enabled because I used emerge to get it)
python-cjson 1.0.5
Python 2.5.1


simplejson results:
% time ./simplejsonbench.py
./simplejsonbench.py 38.97s user 0.00s system 99% cpu 38.968 total
% time ./simplejsonbench.py
./simplejsonbench.py 39.17s user 0.01s system 99% cpu 39.187 total


cjson results:
% time ./cjsonbenchmark.py
./cjsonbenchmark.py 0.77s user 0.01s system 99% cpu 0.784 total
% time ./cjsonbenchmark.py
./cjsonbenchmark.py 0.77s user 0.03s system 100% cpu 0.794 total


simplejson average: 39.07s
cjson average: 0.77s

verdict: cjson 50.74x faster in this benchmark

I’d run more, but I don’t really think I have to. Simplejson actually seems to have a more robust interface than cjson, and is more portable because it doesn’t require a C compiler.

The cjson website has some other numbers that measure actual measured throughput of cjson vs. simplejson. The versions of simplejson is older, and it states there’s no C extensions compiled in.

Tags: benchmarks, cjson, json, Python, simplejson

This entry was posted on Friday, January 25th, 2008 at 10:28 am and is filed under Uncategorized. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.