Ultrajson -- Python's fast JSON Decoder

Lai Yonghao (http://laiyonghao.com)

Note: Not sentence-by-Sentence Translation, if you are confused, Parameter Details of the original (http://pushingtheweb.com/2011/03/ultra-fast-json-encoding-decoding-python ).

In ESN, we use JSON in large quantities. The Python 2.6 comes with a slow speed, so we will use simplejson later. Today, JavaScript and JSON are more widely used, so I think it may be a field worthy of my use of C for optimization, and it can be noted that I am not old.

Ultrajson

I spent a weekend studying JSON encoding and called the result ultrajson. From the preliminary benchmark test I have conducted, he is the fastest among all the JSON codecs I can find (if not, I can optimize it faster !).

Python binding

In fact, ultrajson is not only used in a specific language, but for convenience of use by my colleagues, I implemented its Python binding, which is called the ujson module.
Update: ultrajson can be found in pypi now. The package name is ujson. you can install it easily with easy_install or Pip, you know.

Current Benchmark Test

64-bit Linux Benchmark Test

Python 2.6.6 (r266: 84292, Sep 15 2010, 16:22:56)
OS Version: Ubuntu 10.10
System type: x64-based PC
Processor: Intel (r) core (TM) i5-2300 CPU @ 2.80 GHz
Total physical memory: 4096 MB

Array with 256 UTF-8 strings:
Ujson encode: 2874.54652 CILS/sec
Simplejson encode: 1539.47999 CILS/sec
Cjson encode: 132.33571 CILS/sec

Ujson Decode: 2072.09417 CILS/sec
Cjson Decode: 991.20903 CILS/sec
Simplejson Decode: 310.75309 CILS/sec

Medium complex object:
Ujson encode: 19001.01929 CILS/sec
Simplejson encode: 3512.29205 CILS/sec
Cjson encode: 3063.69959 CILS/sec

Ujson Decode: 12791.80993 CILS/sec
Cjson Decode: 8288.32916 CILS/sec
Simplejson Decode: 6640.22169 CILS/sec

Array with 256 strings:
Ujson encode: 40161.78453 CILS/sec
Simplejson encode: 19301.40779 CILS/sec
Cjson encode: 12337.13166 CILS/sec

Ujson Decode: 36944.81317 CILS/sec
Cjson Decode: 30187.40167 CILS/sec
Simplejson Decode: 25105.56562 CILS/sec

Array with 256 doubles:
Ujson encode: 6054.71950 CILS/sec
Simplejson encode: 2912.44353 CILS/sec
Cjson encode: 3539.51228 CILS/sec

Ujson Decode: 27794.29735 CILS/sec
Cjson Decode: 14892.38775 CILS/sec
Simplejson Decode: 14879.00070 CILS/sec

Array with 256 true values:
Ujson encode: 168086.95325 CILS/sec
Simplejson encode: 49348.93309 CILS/sec
Cjson encode: 67392.90623 CILS/sec

Ujson Decode: 139359.25968 CILS/sec
Cjson Decode: 82552.26652 CILS/sec
Simplejson Decode: 114998.51396 CILS/sec

Array with 256 dict {string, int} pairs:
Ujson encode: 24125.68837 CILS/sec
Simplejson encode: 5751.74871 CILS/sec
Cjson encode: 4735.65147 CILS/sec

Ujson Decode: 17176.70493 CILS/sec
Cjson Decode: 13420.93963 CILS/sec
Simplejson Decode: 9854.27352 CILS/sec

Dict with 256 arrays with 256 dict {string, int} pairs:
Ujson encode: 86.52449 CILS/sec
Simplejson encode: 17.46117 CILS/sec
Cjson encode: 18.31323 CILS/sec

Ujson Decode: 49.54660 CILS/sec
Cjson Decode: 38.34094 CILS/sec
Simplejson Decode: 28.18035 CILS/sec

More on GitHub

I'm looking forward to seeing someone using this project. What's better is contribution.CodeSo come to my GitHub repository (https://github.com/jskorpan/ultrajson.
It would be cool to bind it to more languages!