OvermindDL1 / benchmark-large-json-parsing

Benchmark large json parsing

A few crude benchmarks for Elixir, Golang, C++, and Ruby

Processor Intel(R) Core(TM) i7-5820K CPU @ 3.30GHz
OS: Ubuntu 16.04

10mb.json

Golang [jq]

$ time jq '.' data/10mb.json > /dev/null
real    0m0.567s
user    0m0.551s
sys     0m0.016s

Elixir

Time taken [Poison]: 1218.831ms
Time taken [Jason]: 508.461ms

Ruby

time ruby app.rb

real    0m0.220s
user    0m0.203s
sys     0m0.017s

100mb.json (https://github.com/zemirco/sf-city-lots-json/blob/master/citylots.json)

Golang [jq]

$ time jq '.' data/citylots.json > /dev/null

real    0m14.436s
user    0m13.992s
sys     0m0.420s

Elixir

  Time taken [Poison]: 32_640.87ms
  Time taken [Jason]: 11_602.128ms

Ruby

  $ time ruby app.rb

  real    0m4.738s
  user    0m4.498s
  sys     0m0.240s

Building instructions:

C++

RapidJson

Just run the build.sh script from within the cpp/rapidjson directory and it will output two binaries into the _builds directory, one for sax-style parsing, the other for structure/document style parsing.

To run it the C++ programs will accept it via stdin, so don't pass it in as an argument, but instead pipe it into to it like via time _builds/rapidjson-structure < ../../data/citylots.json or so.